diff --git a/.github/copilot-instructions.md b/.github/copilot-instructions.md
index d5f96553d..46a066d2b 100644
--- a/.github/copilot-instructions.md
+++ b/.github/copilot-instructions.md
@@ -160,6 +160,12 @@ Notes:
   `docs/dev/package-structure/short.md` automatically — never edit those
   by hand. Don't review auto-fixes; accept and move on. Then
   `pixi run check` until clean.
+- When a check command needs saved output for analysis, capture the log
+  and preserve the command exit code with a zsh-safe variable name:
+  `pixi run check > /tmp/easydiffraction-check.log 2>&1; check_exit_code=$?; tail -n 200 /tmp/easydiffraction-check.log; exit $check_exit_code`.
+  Never assign to `status` in zsh; it is readonly. Use task-specific
+  names such as `check_exit_code`, `unit_tests_exit_code`, or
+  `script_tests_exit_code`.
 - Open issues / design questions / planned improvements live in
   `docs/dev/issues/open.md` (priority-ordered). On resolution, move to
   `docs/dev/issues/closed.md` and update the relevant ADR or
@@ -204,6 +210,9 @@ When asked to create a plan:
   files likely to change, decisions already made, open questions,
   verification commands for Phase 2, and a short suggested commit
   message or branch name when useful.
+- Verification commands in plans must include the zsh-safe log-capture
+  pattern from **Workflow** whenever saved output is needed for later
+  analysis.
 - Before saving a plan, verify that referenced files, directories,
   scripts, and task names exist locally when that is practical. If a
   referenced tool is optional or missing, include an available fallback.
diff --git a/.github/workflows/docs.yml b/.github/workflows/docs.yml
index def30827c..ea6bdad11 100644
--- a/.github/workflows/docs.yml
+++ b/.github/workflows/docs.yml
@@ -119,7 +119,7 @@ jobs:
       # Uses multiple cores for parallel execution to speed up the process.
       - name: Run notebooks
         # if: false # Temporarily disabled to speed up the docs build
-        run: pixi run notebook-exec
+        run: pixi run notebook-exec-ci
 
       # Build the static files for the documentation site for local inspection
       # Input: docs/ directory containing the Markdown files
diff --git a/docs/dev/adrs/accepted/analysis-cif-fit-state.md b/docs/dev/adrs/accepted/analysis-cif-fit-state.md
new file mode 100644
index 000000000..a38f0e7c7
--- /dev/null
+++ b/docs/dev/adrs/accepted/analysis-cif-fit-state.md
@@ -0,0 +1,193 @@
+# ADR: Analysis CIF Fit State
+
+## Status
+
+Accepted current design.
+
+## Date
+
+2026-05-18
+
+## Group
+
+Analysis and fitting.
+
+## Context
+
+`analysis/analysis.cif` already persists analysis configuration such as
+`_fitting.minimizer_type`, `_fitting.mode_type`, aliases, constraints,
+and active fit-mode settings. That configuration alone is not enough to
+reopen a saved project and continue the same fit-result, plotting, and
+command-line workflow.
+
+Analysis-owned fit state needs to persist:
+
+- fit bounds and bound provenance
+- pre-fit scalar snapshots for recovery workflows
+- compact status metadata for the latest saved fit projection
+- deterministic correlation summaries
+- Bayesian summary metadata and manifests for bulk array sidecars
+- plot-ready Bayesian caches so restored posterior displays do not need
+  to recompute on first use
+
+Committed model parameter values and uncertainties already persist in
+structure and experiment CIF files through the accepted free-flag CIF
+encoding. Those committed values must remain the source of truth for the
+current model state.
+
+The accepted runtime fit-results ADR keeps backend runtime objects
+runtime-only unless a narrower persistence ADR defines a saved
+projection. This ADR defines that narrower saved projection.
+
+## Decision
+
+Persist analysis-owned fit state as explicit sibling categories in
+`analysis/analysis.cif`, with large Bayesian arrays stored in
+`analysis/results.h5`.
+
+Do not add a dedicated `_fit_state` category or
+`_fit_state.schema_version`. Persisted fit state is detected from
+`_fit_result` and the related fit-state categories.
+
+### Common fit-state categories
+
+Persist these common categories for any saved fit projection:
+
+- `_fit_parameter`
+- `_fit_result`
+- `_fit_parameter_correlation`
+
+`_fit_parameter` stores analysis-owned per-parameter fit controls and
+pre-fit scalar snapshots:
+
+- `param_unique_name`
+- `fit_min`
+- `fit_max`
+- `fit_bounds_uncertainty_multiplier`
+- `start_value`
+- `start_uncertainty`
+
+`_fit_result` stores the latest saved fit header:
+
+- `result_kind`
+- `success`
+- `message`
+- `iterations`
+- `fitting_time`
+- `reduced_chi_square`
+
+`_fit_parameter_correlation` stores pairwise deterministic or posterior
+correlation summaries keyed by a persisted `id`. Only unique parameter
+pairs are stored.
+
+### Deterministic fit projection
+
+Deterministic fits persist `_deterministic_result` in addition to the
+common categories above.
+
+`_deterministic_result` stores compact optimizer metadata and counts:
+
+- `optimizer_name`
+- `method_name`
+- `objective_name`
+- `objective_value`
+- `n_data_points`
+- `n_parameters`
+- `n_free_parameters`
+- `degrees_of_freedom`
+- `covariance_available`
+- `correlation_available`
+
+Do not persist a `_deterministic_parameter_result` category. Final
+deterministic parameter values and uncertainties already persist in the
+model CIF files, and restored deterministic ordering comes from
+`_fit_parameter`.
+
+### Bayesian fit projection
+
+Bayesian fits persist these additional categories:
+
+- `_bayesian_result`
+- `_bayesian_sampler`
+- `_bayesian_convergence`
+- `_bayesian_parameter_posterior`
+- `_bayesian_distribution_cache`
+- `_bayesian_pair_cache`
+- `_bayesian_predictive_dataset`
+
+`_bayesian_result` stores the saved Bayesian header and sidecar flags,
+including `sidecar_file`, `has_posterior_samples`,
+`has_distribution_cache`, `has_pair_cache`, and
+`has_posterior_predictive`.
+
+`_bayesian_sampler` stores the resolved sampler settings used for the
+run. `parallel` persists the resolved non-negative worker count as an
+integer.
+
+`_bayesian_convergence` stores convergence metadata and posterior array
+shape counts.
+
+`_bayesian_parameter_posterior` stores one summary row per sampled
+parameter, including credible intervals, uncertainty, ESS, and R-hat.
+Its row order defines the saved posterior parameter order.
+
+`_bayesian_distribution_cache`, `_bayesian_pair_cache`, and
+`_bayesian_predictive_dataset` store manifest rows for plot-ready
+posterior caches. Distribution and predictive caches are persisted for
+any Bayesian fit with posterior samples, including single-parameter
+fits. Pair caches and posterior correlation summaries are only persisted
+when more than one parameter was sampled.
+
+`parameter.posterior` is not part of this accepted design. This ADR
+persists analysis-level posterior summaries and caches only. Any future
+parameter-level posterior API remains a separate decision.
+
+### Bayesian sidecar
+
+Persist large Bayesian arrays in `analysis/results.h5` using `h5py`.
+This includes canonical posterior arrays and any saved distribution,
+pair, and predictive cache arrays referenced by the CIF manifests.
+
+The persisted `sidecar_file` value is a local file name only. It must
+resolve to a basename inside the project `analysis/` directory. Absolute
+paths and traversal paths are rejected and fall back to `results.h5`.
+
+If the sidecar is missing on load, summary rows in
+`analysis/analysis.cif` still restore fit tables and metadata. Features
+that require missing bulk arrays must warn clearly instead of failing
+silently.
+
+### Save and restore behavior
+
+After a fit completes, project save writes the fit-state projection
+before the project is considered fully persisted. For Bayesian fits,
+that includes the prepared summaries and saved plot caches used by
+posterior displays.
+
+Load order is:
+
+1. standard analysis configuration
+2. common fit-state categories
+3. deterministic or Bayesian fit-specific categories according to
+   `_fit_result.result_kind`
+4. Bayesian sidecar arrays when a Bayesian sidecar is expected
+
+Persist backend runtime objects, optimizer instances, and raw driver
+payloads nowhere in this design.
+
+## Consequences
+
+Saved projects reopen with enough fit-state context to display the last
+saved result and rerun fits without rebuilding analysis-owned bounds by
+hand.
+
+Deterministic persistence stays compact because committed parameter
+values remain in the model CIF files instead of being duplicated in a
+second deterministic per-parameter result loop.
+
+Bayesian persistence spans CIF metadata and an HDF5 sidecar, so save and
+load must validate consistency between manifest rows and bulk datasets.
+
+The accepted runtime fit-results ADR should now be read as runtime-only
+except where this narrower projection explicitly persists fit-state
+metadata, summaries, and cache arrays.
diff --git a/docs/dev/adrs/suggestions/loop-category-key-identity.md b/docs/dev/adrs/accepted/loop-category-key-identity.md
similarity index 78%
rename from docs/dev/adrs/suggestions/loop-category-key-identity.md
rename to docs/dev/adrs/accepted/loop-category-key-identity.md
index f55ef4a89..4334b4e13 100644
--- a/docs/dev/adrs/suggestions/loop-category-key-identity.md
+++ b/docs/dev/adrs/accepted/loop-category-key-identity.md
@@ -1,6 +1,6 @@
 # ADR: Loop Category Keys and Identity Naming
 
-**Status:** Proposed  
+**Status:** Accepted  
 **Date:** 2026-05-18
 
 ## Context
@@ -10,7 +10,7 @@ CIF dictionaries can declare the key column for a loop category through
 use domain-specific identity tags such as `_atom_site.label`, while
 other loop categories may use an explicit `id` tag.
 
-EasyDiffraction currently models the same runtime concept with
+EasyDiffraction models the same runtime concept with
 `item._identity.category_entry_name`. `CategoryCollection` uses this
 value as the collection key, and category items use it in their
 `unique_name` path:
@@ -19,23 +19,24 @@ value as the collection key, and category items use it in their
 <datablock>.<category_code>.<category_entry_name>
 ```
 
-The design question is whether the current `category_entry_name`
-approach is enough, and how closely Python-facing identity names should
-follow CIF key tags.
+The design question was whether the `category_entry_name` approach was
+enough, and how closely Python-facing identity names should follow CIF
+key tags.
 
 ## Assessment
 
-The current approach is directionally good. It gives every loop item a
-stable collection key without hard-coding the key field into
+The `category_entry_name` approach is directionally good. It gives every
+loop item a stable collection key without hard-coding the key field into
 `CategoryCollection`, and most current loop categories derive that key
 from the same field that is serialized into CIF.
 
-It is not explicit enough yet. The key field is encoded as a lambda on
-each item, not as declarative metadata on the category or descriptor.
-Nothing validates that the runtime key corresponds to a serialized CIF
-field. The main visible example is `Constraint`: the current collection
-key is derived from the left-hand side of `_constraint.expression`, but
-no separate `_constraint.id` field is persisted.
+Before this decision it was not explicit enough. The key field was
+encoded as a lambda on each item, not as declarative metadata on the
+category or descriptor. Nothing validated that the runtime key
+corresponded to a serialized CIF field. The main visible example was
+`Constraint`: the collection key was derived from the left-hand side of
+`_constraint.expression`, but no separate `_constraint.id` field was
+persisted.
 
 ## Decision
 
@@ -52,14 +53,19 @@ Keep `category_entry_name` as the runtime analogue of CIF
    identity. Use `label` or `*_id` when the value has clearer domain
    meaning.
 
-The `constraint` category should add an explicit `id` field and use it
-as the collection key:
+The `constraint` category has an explicit `id` field and uses it as the
+collection key:
 
 ```text
 analysis.constraints[id].id -> _constraint.id
 ```
 
-The existing `lhs_alias` and `rhs_expr` properties should remain derived
+`Constraints.create()` accepts an optional `id` argument. A
+user-provided `id` is used whenever it is not `None`; otherwise the
+method derives the default `id` from `lhs_alias`, preserving the old
+`create(expression=...)` behavior.
+
+The existing `lhs_alias` and `rhs_expr` properties remain derived
 helpers from `_constraint.expression`, not the row identity.
 
 This argues against a blanket Python API change to use `id` everywhere.
@@ -74,22 +80,22 @@ Rows are sorted by the chosen Python key style: `id`, then `*_id`, then
 identifier-like fields that are not collection keys are listed in the
 next section.
 
-| Python key      | Area       | Collection class                            | Category code            | CIF key tag                  | Source                      | Decision                                                                                             |
-| --------------- | ---------- | ------------------------------------------- | ------------------------ | ---------------------------- | --------------------------- | ---------------------------------------------------------------------------------------------------- |
-| `id`            | Analysis   | `Constraints`                               | `constraint`             | `_constraint.id`             | Custom category             | Add this key. Current implementation derives the key from the left side of `_constraint.expression`. |
-| `id`            | Analysis   | `SequentialFitExtractCollection`            | `sequential_fit_extract` | `_sequential_fit_extract.id` | Custom category             | Keep. It is an explicit row identifier for extraction rules.                                         |
-| `id`            | Experiment | `LineSegmentBackground`                     | `background`             | `_pd_background.id`          | Powder CIF-style category   | Keep. The row identity is opaque and already serialized.                                             |
-| `id`            | Experiment | `ChebyshevPolynomialBackground`             | `background`             | `_pd_background.id`          | Powder CIF-style category   | Keep. The row identity is opaque and shared with other background variants.                          |
-| `id`            | Experiment | `LinkedPhases`                              | `linked_phases`          | `_pd_phase_block.id`         | Powder CIF-style category   | Keep. Consider `phase_id` only if the public API later standardizes foreign-key names.               |
-| `id`            | Experiment | `ExcludedRegions`                           | `excluded_regions`       | `_excluded_region.id`        | Custom category             | Keep. It is a simple custom loop row identifier.                                                     |
-| `id`            | Experiment | `ReflnData`                                 | `refln`                  | `_refln.id`                  | CIF-style category          | Keep for the current reflection table shape.                                                         |
-| `id`            | Experiment | `PowderCwlReflnData` / `PowderTofReflnData` | `refln`                  | `_refln.id`                  | CIF-style category          | Keep; `phase_id` remains a separate field, not the row key.                                          |
-| `experiment_id` | Analysis   | `JointFitCollection`                        | `joint_fit`              | `_joint_fit.experiment_id`   | Custom category             | Keep. The key is a reference to an experiment, so `id` alone would lose context.                     |
-| `point_id`      | Experiment | `PdCwlData` / `PdTofData`                   | `pd_data`                | `_pd_data.point_id`          | Powder CIF-style category   | Keep. It is clearer than `id` for dense measured/calculated data points.                             |
-| `point_id`      | Experiment | `TotalData`                                 | `total_data`             | `_pd_data.point_id`          | Current powder-data mapping | Keep. Revisit the CIF tag only when total-scattering-specific CIF tags are introduced.               |
-| `label`         | Analysis   | `Aliases`                                   | `alias`                  | `_alias.label`               | Custom category             | Keep. It is the user-visible symbol referenced by expressions, not an opaque row id.                 |
-| `label`         | Structure  | `AtomSites`                                 | `atom_site`              | `_atom_site.label`           | CIF core category           | Keep. This is a well-known crystallographic identity field.                                          |
-| `label`         | Structure  | `AtomSiteAnisoCollection`                   | `atom_site_aniso`        | `_atom_site_aniso.label`     | CIF core category           | Keep. It intentionally matches and references the atom-site label.                                   |
+| Python key      | Area       | Collection class                            | Category code            | CIF key tag                  | Source                      | Decision                                                                                               |
+| --------------- | ---------- | ------------------------------------------- | ------------------------ | ---------------------------- | --------------------------- | ------------------------------------------------------------------------------------------------------ |
+| `id`            | Analysis   | `Constraints`                               | `constraint`             | `_constraint.id`             | Custom category             | Add this key. Previous implementations derived the key from the left side of `_constraint.expression`. |
+| `id`            | Analysis   | `SequentialFitExtractCollection`            | `sequential_fit_extract` | `_sequential_fit_extract.id` | Custom category             | Keep. It is an explicit row identifier for extraction rules.                                           |
+| `id`            | Experiment | `LineSegmentBackground`                     | `background`             | `_pd_background.id`          | Powder CIF-style category   | Keep. The row identity is opaque and already serialized.                                               |
+| `id`            | Experiment | `ChebyshevPolynomialBackground`             | `background`             | `_pd_background.id`          | Powder CIF-style category   | Keep. The row identity is opaque and shared with other background variants.                            |
+| `id`            | Experiment | `LinkedPhases`                              | `linked_phases`          | `_pd_phase_block.id`         | Powder CIF-style category   | Keep. Consider `phase_id` only if the public API later standardizes foreign-key names.                 |
+| `id`            | Experiment | `ExcludedRegions`                           | `excluded_regions`       | `_excluded_region.id`        | Custom category             | Keep. It is a simple custom loop row identifier.                                                       |
+| `id`            | Experiment | `ReflnData`                                 | `refln`                  | `_refln.id`                  | CIF-style category          | Keep for the current reflection table shape.                                                           |
+| `id`            | Experiment | `PowderCwlReflnData` / `PowderTofReflnData` | `refln`                  | `_refln.id`                  | CIF-style category          | Keep; `phase_id` remains a separate field, not the row key.                                            |
+| `experiment_id` | Analysis   | `JointFitCollection`                        | `joint_fit`              | `_joint_fit.experiment_id`   | Custom category             | Keep. The key is a reference to an experiment, so `id` alone would lose context.                       |
+| `point_id`      | Experiment | `PdCwlData` / `PdTofData`                   | `pd_data`                | `_pd_data.point_id`          | Powder CIF-style category   | Keep. It is clearer than `id` for dense measured/calculated data points.                               |
+| `point_id`      | Experiment | `TotalData`                                 | `total_data`             | `_pd_data.point_id`          | Current powder-data mapping | Keep. Revisit the CIF tag only when total-scattering-specific CIF tags are introduced.                 |
+| `label`         | Analysis   | `Aliases`                                   | `alias`                  | `_alias.label`               | Custom category             | Keep. It is the user-visible symbol referenced by expressions, not an opaque row id.                   |
+| `label`         | Structure  | `AtomSites`                                 | `atom_site`              | `_atom_site.label`           | CIF core category           | Keep. This is a well-known crystallographic identity field.                                            |
+| `label`         | Structure  | `AtomSiteAnisoCollection`                   | `atom_site_aniso`        | `_atom_site_aniso.label`     | CIF core category           | Keep. It intentionally matches and references the atom-site label.                                     |
 
 ## Non-Key Identity And Reference Fields
 
@@ -123,22 +129,21 @@ Jupyter, CLI output, and saved CIF files.
 
 ## Implementation Notes
 
-The current `category_entry_name` mechanism can stay, but it should be
-made easier to audit. The implementation now uses class-level
-`_category_code` and `_category_entry_name` declarations on concrete
-`CategoryItem` subclasses. `CategoryItem` resolves the declared
-`_category_entry_name` lazily from the named public attribute, and
-`Identity` exposes the resolved value through
-`item._identity.category_entry_name`.
+The implementation uses class-level `_category_code` and
+`_category_entry_name` declarations on concrete `CategoryItem`
+subclasses. `CategoryItem` resolves the declared `_category_entry_name`
+lazily from the named public attribute, and `Identity` exposes the
+resolved value through `item._identity.category_entry_name`.
 
-For constraints, add a descriptor-backed `id` property serialized as
-`_constraint.id`, and change `category_entry_name` to resolve from that
-descriptor. Keep `_constraint.expression` for the full equation. Keep
-`lhs_alias` and `rhs_expr` as derived convenience properties.
+For constraints, a descriptor-backed `id` property is serialized as
+`_constraint.id`, and `category_entry_name` resolves from that
+descriptor. `_constraint.expression` stores the full equation.
+`lhs_alias` and `rhs_expr` remain derived convenience properties.
 
 When reading older CIF files that only contain `_constraint.expression`,
-derive a deterministic fallback `id` from the old `lhs_alias` key after
-row values are loaded, then write `_constraint.id` on the next save.
+the loader derives a deterministic fallback `id` from the old
+`lhs_alias` key after row values are loaded, then writes
+`_constraint.id` on the next save.
 
 ## Consequences
 
diff --git a/docs/dev/adrs/accepted/parameter-correlation-persistence.md b/docs/dev/adrs/accepted/parameter-correlation-persistence.md
new file mode 100644
index 000000000..9b51c0362
--- /dev/null
+++ b/docs/dev/adrs/accepted/parameter-correlation-persistence.md
@@ -0,0 +1,104 @@
+# ADR: Parameter Correlation Persistence
+
+## Status
+
+Accepted current design.
+
+## Date
+
+2026-05-19
+
+## Group
+
+Analysis and fitting.
+
+## Context
+
+`plot_param_correlations()` can visualize either deterministic parameter
+correlations derived from engine covariance or Bayesian correlations
+derived from posterior samples.
+
+Reloaded projects still need correlation heatmaps even when the raw
+runtime covariance or posterior arrays are unavailable. The broader
+fit-state layout is defined in `analysis-cif-fit-state.md`; this ADR
+records the narrower persisted correlation-summary projection within
+that accepted design.
+
+Correlation data is analysis-owned derived state, not model state. It
+therefore belongs in `analysis/analysis.cif`, not in structure or
+experiment CIF files.
+
+## Decision
+
+Persist pairwise parameter correlations in `_fit_parameter_correlation`
+rows inside `analysis/analysis.cif`.
+
+### Correlation summary schema
+
+Store one row per unique parameter pair with these fields:
+
+- `id`
+- `source_kind`
+- `param_unique_name_i`
+- `param_unique_name_j`
+- `correlation`
+
+Example:
+
+```cif
+loop_
+_fit_parameter_correlation.id
+_fit_parameter_correlation.source_kind
+_fit_parameter_correlation.param_unique_name_i
+_fit_parameter_correlation.param_unique_name_j
+_fit_parameter_correlation.correlation
+1 posterior cosio.atom_site.Co1.adp_iso cosio.atom_site.Co2.adp_iso 0.87
+```
+
+Normalize each row to the upper triangle excluding the diagonal.
+`param_unique_name_i` and `param_unique_name_j` use a stable ordering so
+only one unordered pair is stored. The diagonal is omitted because it is
+always `1.0` and can be rebuilt on load.
+
+Use the same loop for deterministic and Bayesian projections. The source
+is carried by `source_kind`, currently `deterministic` or `posterior`.
+
+### Summary-only role
+
+`_fit_parameter_correlation` is a persisted summary. It does not replace
+posterior samples, posterior pair densities, or covariance matrices.
+
+This summary is enough to restore correlation heatmaps. It is not enough
+to restore richer pair-plot density surfaces or covariance-specific
+workflows.
+
+### Restore behavior
+
+On load, restore `_fit_parameter_correlation` rows into an analysis-
+owned correlation collection.
+
+When runtime covariance or posterior samples are unavailable,
+`project.display.plotter.plot_param_correlations()` may rebuild a square
+correlation matrix from the persisted rows and the restored fit-result
+parameter ordering.
+
+### Relation to the broader fit-state ADR
+
+`analysis-cif-fit-state.md` remains the source of truth for the full
+fit-state projection, save/load ordering, and Bayesian sidecar layout.
+This ADR records the implemented correlation-specific piece of that
+accepted design.
+
+## Consequences
+
+- Deterministic and Bayesian correlation summaries survive reload.
+- Correlation heatmaps no longer depend entirely on runtime-only data.
+- The schema stays compact and fit-type-agnostic.
+- Posterior pair plots and covariance-specific workflows still need
+  richer runtime or persisted data.
+
+## Deferred Work
+
+- optional storage of covariance matrices in addition to correlations
+- multiple named correlation sources for the same saved project
+- full correlation restoration for pair-plot density surfaces
diff --git a/docs/dev/adrs/accepted/project-facade-and-persistence.md b/docs/dev/adrs/accepted/project-facade-and-persistence.md
index 67492d8d4..3b922f423 100644
--- a/docs/dev/adrs/accepted/project-facade-and-persistence.md
+++ b/docs/dev/adrs/accepted/project-facade-and-persistence.md
@@ -14,16 +14,22 @@ Persistence.
 
 ## Context
 
-`Project` is the current top-level user facade. It owns project
-metadata, structures, experiments, rendering preferences, display
-helpers, analysis, summaries, verbosity, and save/load behavior.
+`Project` is the top-level user facade. It owns project metadata,
+structures, experiments, rendering preferences, display helpers,
+analysis, summaries, verbosity, and save/load behavior.
+
+A later proposal considered renaming this facade to `Workspace` so that
+`project` could be reserved for the scientific project information
+category. The final naming decision is to keep `Project` as the public
+root because it matches scientific user language and the saved project
+container.
 
 The persisted project directory needs to separate real CIF datablocks
 from singleton project sections.
 
 ## Decision
 
-Use `Project` as the current top-level facade and persist projects as a
+Use `Project` as the top-level facade and persist projects as a
 directory of CIF files:
 
 ```text
@@ -40,8 +46,35 @@ Real structures and experiments serialize as `data_<id>` datablocks.
 Singleton sections such as project configuration, analysis, and summary
 serialize without fake `data_` headers.
 
+Keep project information available as `project.info`. The Python name
+avoids a confusing `project.project` access path, while the persisted
+CIF category remains the semantic `_project.*` category:
+
+```cif
+_project.id
+_project.title
+_project.description
+_project.created
+_project.last_modified
+```
+
+Do not introduce `_meta.*` tags for project information. The category is
+scientific project information, not generic metadata. Any future change
+from `_project.*` to another category name must be a separate explicit
+persistence decision.
+
+Keep `project.cif` as the primary singleton project configuration file
+while `Project` remains the root facade. Do not rename it to
+`workspace.cif` as a side effect of category cleanup.
+
+The saved project directory path is runtime file-I/O state, not a
+serialized project-information field. If the path is exposed in Python,
+it must not emit a `_project.path` CIF item.
+
 ## Consequences
 
 The saved layout mirrors the current object graph while preserving the
-semantic difference between real datablocks and singleton sections. A
-proposed `Workspace` rename is tracked separately as an ADR suggestion.
+semantic difference between real datablocks and singleton sections. The
+`Workspace` rename proposal is rejected; ADR examples should continue to
+use `Project`, `project.info`, and `project.cif` unless a later accepted
+ADR changes a narrower part of this design.
diff --git a/docs/dev/adrs/index.md b/docs/dev/adrs/index.md
index ed0940bf3..50a287db9 100644
--- a/docs/dev/adrs/index.md
+++ b/docs/dev/adrs/index.md
@@ -13,34 +13,34 @@ folders.
 
 ## ADR Index
 
-| Group                | Status     | Title                                                         | Short description                                                                                                  | Link                                                                                         |
-| -------------------- | ---------- | ------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------ | -------------------------------------------------------------------------------------------- |
-| Analysis and fitting | Accepted   | Fit Mode Categories and Fit Execution API                     | Splits fitting configuration from execution and defines active sibling fit-mode categories.                        | [`fit-mode-categories.md`](accepted/fit-mode-categories.md)                                  |
-| Analysis and fitting | Accepted   | Runtime Fit Results                                           | Keeps full fit outputs runtime-only in the current design unless a narrower persistence ADR is accepted.           | [`runtime-fit-results.md`](accepted/runtime-fit-results.md)                                  |
-| Analysis and fitting | Suggestion | Analysis CIF Fit State                                        | Proposes a persisted scalar projection of fit state in `analysis.cif`.                                             | [`analysis-cif-fit-state.md`](suggestions/analysis-cif-fit-state.md)                         |
-| Analysis and fitting | Suggestion | Parameter Correlation Persistence                             | Proposes persisting deterministic and posterior correlation summaries.                                             | [`parameter-correlation-persistence.md`](suggestions/parameter-correlation-persistence.md)   |
-| Analysis and fitting | Suggestion | Parameter-Level Posterior Projection and Bayesian Persistence | Proposes saved Bayesian summaries and canonical posterior storage.                                                 | [`parameter-posterior-summary.md`](suggestions/parameter-posterior-summary.md)               |
-| Analysis and fitting | Suggestion | Undo Fit                                                      | Proposes an analysis-owned rollback operation for the latest pre-fit scalar state.                                 | [`undo-fit.md`](suggestions/undo-fit.md)                                                     |
-| Core model           | Accepted   | Category Owners and Real Datablocks                           | Introduces `CategoryOwner` so singleton sections do not pretend to be real CIF datablocks.                         | [`category-owner-sections.md`](accepted/category-owner-sections.md)                          |
-| Core model           | Accepted   | Enum-Backed Closed Value Sets                                 | Requires finite option sets to use `(str, Enum)` classes for validation and dispatch.                              | [`enum-backed-closed-values.md`](accepted/enum-backed-closed-values.md)                      |
-| Core model           | Accepted   | Guarded Public Properties                                     | Uses property setters as the public writability contract for guarded objects.                                      | [`guarded-public-properties.md`](accepted/guarded-public-properties.md)                      |
-| Core model           | Accepted   | Two-Level Category Parameter Access                           | Keeps parameter access to `datablock.category.parameter` or `datablock.collection[id].parameter`.                  | [`category-parameter-access.md`](accepted/category-parameter-access.md)                      |
-| Documentation        | Accepted   | Descriptor Property Docstring Template                        | Makes descriptor metadata the source of truth for public property docstrings and annotations.                      | [`property-docstring-template.md`](accepted/property-docstring-template.md)                  |
-| Documentation        | Accepted   | Development Documentation Structure                           | Defines the `docs/dev` layout for ADRs, issues, plans, package structure, and roadmap.                             | [`development-docs-structure.md`](accepted/development-docs-structure.md)                    |
-| Documentation        | Accepted   | Help Method Discoverability                                   | Requires primary public objects and facades to expose consistent `help()` output.                                  | [`help-discoverability.md`](accepted/help-discoverability.md)                                |
-| Documentation        | Accepted   | Notebook Generation Source of Truth                           | Treats tutorial `.py` files as editable sources and notebooks as generated artifacts.                              | [`notebook-generation.md`](accepted/notebook-generation.md)                                  |
-| Experiment model     | Accepted   | Immutable Experiment Type                                     | Makes experiment type axes creation-time state rather than mutable runtime state.                                  | [`immutable-experiment-type.md`](accepted/immutable-experiment-type.md)                      |
-| Factories            | Accepted   | Factory Contracts and Metadata                                | Standardizes factory construction, metadata, compatibility, and registration behavior.                             | [`factory-contracts.md`](accepted/factory-contracts.md)                                      |
-| Naming               | Accepted   | Factory Tag Naming                                            | Defines canonical factory tag style and standard abbreviations.                                                    | [`factory-tag-naming.md`](accepted/factory-tag-naming.md)                                    |
-| Persistence          | Accepted   | Free-Flag CIF Encoding                                        | Encodes fit free/fixed state through CIF uncertainty syntax instead of a separate free list.                       | [`free-flag-cif-encoding.md`](accepted/free-flag-cif-encoding.md)                            |
-| Persistence          | Accepted   | Project Facade and Persistence Layout                         | Documents the current `Project` facade and saved directory layout.                                                 | [`project-facade-and-persistence.md`](accepted/project-facade-and-persistence.md)            |
-| Persistence          | Suggestion | Loop Category Keys and Identity Naming                        | Documents current loop collection keys and proposes naming rules aligned with CIF category keys.                   | [`loop-category-key-identity.md`](suggestions/loop-category-key-identity.md)                 |
-| Persistence          | Suggestion | Python and CIF Category Correspondence                        | Compares current Python paths and CIF tags, then proposes scoped one-to-one mapping for project-level categories.  | [`python-cif-category-correspondence.md`](suggestions/python-cif-category-correspondence.md) |
-| Quality              | Accepted   | Lint Complexity Thresholds                                    | Treats ruff PLR complexity limits as design guardrails that should not be bypassed.                                | [`lint-complexity-thresholds.md`](accepted/lint-complexity-thresholds.md)                    |
-| Quality              | Accepted   | Test Strategy                                                 | Defines layered unit, functional, integration, script, and notebook testing.                                       | [`test-strategy.md`](accepted/test-strategy.md)                                              |
-| Structure model      | Accepted   | Type-Neutral ADP Parameters                                   | Keeps ADP parameter object identities stable across B/U and iso/ani switches.                                      | [`type-neutral-adp-parameters.md`](accepted/type-neutral-adp-parameters.md)                  |
-| User-facing API      | Accepted   | Display UX Facade                                             | Defines `project.display` and `project.rendering` responsibilities and display method names.                       | [`display-ux.md`](accepted/display-ux.md)                                                    |
-| User-facing API      | Accepted   | Selector Families                                             | Distinguishes backend selectors, switchable-category selectors, and active-sibling selectors.                      | [`selector-families.md`](accepted/selector-families.md)                                      |
-| User-facing API      | Accepted   | String Paths and Live Descriptors                             | Separates persisted field selectors from references to live model parameters.                                      | [`string-paths-and-live-descriptors.md`](accepted/string-paths-and-live-descriptors.md)      |
-| User-facing API      | Accepted   | Switchable Category API                                       | Places multi-type category selectors on the owner and omits public selectors for fixed or single-type categories.  | [`switchable-category-api.md`](accepted/switchable-category-api.md)                          |
-| Workspace model      | Suggestion | Workspace Root and Project Information Category               | Proposes renaming the top-level facade from `Project` to `Workspace` and reserving `project` for project metadata. | [`workspace-root-project-category.md`](suggestions/workspace-root-project-category.md)       |
+| Group                | Status     | Title                                     | Short description                                                                                                 | Link                                                                                         |
+| -------------------- | ---------- | ----------------------------------------- | ----------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------- |
+| Analysis and fitting | Accepted   | Fit Mode Categories and Fit Execution API | Splits fitting configuration from execution and defines active sibling fit-mode categories.                       | [`fit-mode-categories.md`](accepted/fit-mode-categories.md)                                  |
+| Analysis and fitting | Accepted   | Runtime Fit Results                       | Keeps full fit outputs runtime-only in the current design unless a narrower persistence ADR is accepted.          | [`runtime-fit-results.md`](accepted/runtime-fit-results.md)                                  |
+| Analysis and fitting | Accepted   | Analysis CIF Fit State                    | Defines the persisted fit-state projection in `analysis/analysis.cif` and `analysis/results.h5`.                  | [`analysis-cif-fit-state.md`](accepted/analysis-cif-fit-state.md)                            |
+| Analysis and fitting | Accepted   | Parameter Correlation Persistence         | Persists deterministic and posterior correlation summaries in `_fit_parameter_correlation`                        | [`parameter-correlation-persistence.md`](accepted/parameter-correlation-persistence.md)      |
+| Analysis and fitting | Suggestion | Fit Output Files and Data Exports         | Narrows remaining archive/export questions after adopting `results.csv` and `results.h5`.                         | [`fit-output-files-and-data-exports.md`](suggestions/fit-output-files-and-data-exports.md)   |
+| Analysis and fitting | Suggestion | Parameter-Level Posterior Projection      | Narrows the still-open `parameter.posterior` API after analysis-level posterior summaries were accepted.          | [`parameter-posterior-summary.md`](suggestions/parameter-posterior-summary.md)               |
+| Analysis and fitting | Suggestion | Undo Fit                                  | Builds rollback semantics and CLI behavior on already-persisted pre-fit scalar snapshots.                         | [`undo-fit.md`](suggestions/undo-fit.md)                                                     |
+| Core model           | Accepted   | Category Owners and Real Datablocks       | Introduces `CategoryOwner` so singleton sections do not pretend to be real CIF datablocks.                        | [`category-owner-sections.md`](accepted/category-owner-sections.md)                          |
+| Core model           | Accepted   | Enum-Backed Closed Value Sets             | Requires finite option sets to use `(str, Enum)` classes for validation and dispatch.                             | [`enum-backed-closed-values.md`](accepted/enum-backed-closed-values.md)                      |
+| Core model           | Accepted   | Guarded Public Properties                 | Uses property setters as the public writability contract for guarded objects.                                     | [`guarded-public-properties.md`](accepted/guarded-public-properties.md)                      |
+| Core model           | Accepted   | Two-Level Category Parameter Access       | Keeps parameter access to `datablock.category.parameter` or `datablock.collection[id].parameter`.                 | [`category-parameter-access.md`](accepted/category-parameter-access.md)                      |
+| Documentation        | Accepted   | Descriptor Property Docstring Template    | Makes descriptor metadata the source of truth for public property docstrings and annotations.                     | [`property-docstring-template.md`](accepted/property-docstring-template.md)                  |
+| Documentation        | Accepted   | Development Documentation Structure       | Defines the `docs/dev` layout for ADRs, issues, plans, package structure, and roadmap.                            | [`development-docs-structure.md`](accepted/development-docs-structure.md)                    |
+| Documentation        | Accepted   | Help Method Discoverability               | Requires primary public objects and facades to expose consistent `help()` output.                                 | [`help-discoverability.md`](accepted/help-discoverability.md)                                |
+| Documentation        | Accepted   | Notebook Generation Source of Truth       | Treats tutorial `.py` files as editable sources and notebooks as generated artifacts.                             | [`notebook-generation.md`](accepted/notebook-generation.md)                                  |
+| Experiment model     | Accepted   | Immutable Experiment Type                 | Makes experiment type axes creation-time state rather than mutable runtime state.                                 | [`immutable-experiment-type.md`](accepted/immutable-experiment-type.md)                      |
+| Factories            | Accepted   | Factory Contracts and Metadata            | Standardizes factory construction, metadata, compatibility, and registration behavior.                            | [`factory-contracts.md`](accepted/factory-contracts.md)                                      |
+| Naming               | Accepted   | Factory Tag Naming                        | Defines canonical factory tag style and standard abbreviations.                                                   | [`factory-tag-naming.md`](accepted/factory-tag-naming.md)                                    |
+| Persistence          | Accepted   | Free-Flag CIF Encoding                    | Encodes fit free/fixed state through CIF uncertainty syntax instead of a separate free list.                      | [`free-flag-cif-encoding.md`](accepted/free-flag-cif-encoding.md)                            |
+| Persistence          | Accepted   | Loop Category Keys and Identity Naming    | Documents loop collection keys and naming rules aligned with CIF category keys.                                   | [`loop-category-key-identity.md`](accepted/loop-category-key-identity.md)                    |
+| Persistence          | Accepted   | Project Facade and Persistence Layout     | Documents the current `Project` facade and saved directory layout.                                                | [`project-facade-and-persistence.md`](accepted/project-facade-and-persistence.md)            |
+| Persistence          | Suggestion | Python and CIF Category Correspondence    | Compares current Python paths and CIF tags, then proposes scoped one-to-one mapping for project-level categories. | [`python-cif-category-correspondence.md`](suggestions/python-cif-category-correspondence.md) |
+| Quality              | Accepted   | Lint Complexity Thresholds                | Treats ruff PLR complexity limits as design guardrails that should not be bypassed.                               | [`lint-complexity-thresholds.md`](accepted/lint-complexity-thresholds.md)                    |
+| Quality              | Accepted   | Test Strategy                             | Defines layered unit, functional, integration, script, and notebook testing.                                      | [`test-strategy.md`](accepted/test-strategy.md)                                              |
+| Structure model      | Accepted   | Type-Neutral ADP Parameters               | Keeps ADP parameter object identities stable across B/U and iso/ani switches.                                     | [`type-neutral-adp-parameters.md`](accepted/type-neutral-adp-parameters.md)                  |
+| User-facing API      | Accepted   | Display UX Facade                         | Defines `project.display` and `project.rendering` responsibilities and display method names.                      | [`display-ux.md`](accepted/display-ux.md)                                                    |
+| User-facing API      | Accepted   | Selector Families                         | Distinguishes backend selectors, switchable-category selectors, and active-sibling selectors.                     | [`selector-families.md`](accepted/selector-families.md)                                      |
+| User-facing API      | Accepted   | String Paths and Live Descriptors         | Separates persisted field selectors from references to live model parameters.                                     | [`string-paths-and-live-descriptors.md`](accepted/string-paths-and-live-descriptors.md)      |
+| User-facing API      | Accepted   | Switchable Category API                   | Places multi-type category selectors on the owner and omits public selectors for fixed or single-type categories. | [`switchable-category-api.md`](accepted/switchable-category-api.md)                          |
diff --git a/docs/dev/adrs/suggestions/analysis-cif-fit-state.md b/docs/dev/adrs/suggestions/analysis-cif-fit-state.md
deleted file mode 100644
index d471bb117..000000000
--- a/docs/dev/adrs/suggestions/analysis-cif-fit-state.md
+++ /dev/null
@@ -1,220 +0,0 @@
-# ADR: Analysis CIF Fit State
-
-**Status:** Proposed  
-**Date:** 2026-05-13
-
-## Context
-
-`analysis/analysis.cif` currently persists analysis configuration such
-as `_fit.minimizer_type`, `_fit.mode`, aliases, constraints, and
-joint-fit weights. It does not persist analysis-owned fit state such as
-fit bounds, bound provenance, pre-fit scalar snapshots, or latest
-fit-status metadata.
-
-At the same time, parameter CIF serialization already carries the
-committed parameter `value`, the current `free` state, and the current
-`uncertainty` via CIF bracket notation. That data belongs to the model
-and should remain in structure or experiment CIF files.
-
-The missing piece is analysis-owned fit state:
-
-- fit controls that apply to parameters during fitting but are not part
-  of the model itself
-- latest fit-status metadata shown by `display.fit_results()`
-- deterministic and Bayesian fit metadata that should survive project
-  reloads and command-line workflows
-
-This separation matters because:
-
-- Bayesian plotting after reload needs `fit_min`, `fit_max`, and bound
-  provenance even when raw posterior arrays are absent
-- command-line users need a saved pre-fit starting state to recover from
-  a poor minimization run
-- `analysis.fit_results` already changes by fit type, but its persisted
-  projection should have a stable analysis-owned home
-
-The accepted runtime-fit-results ADR describes fit results as
-runtime-only. This ADR proposes a narrower persisted projection of the
-latest fit state, not a direct dump of backend runtime objects.
-
-## Decision
-
-### 1. `analysis/analysis.cif` becomes the home of analysis-owned fit state
-
-The analysis CIF file will persist:
-
-- fit configuration
-- aliases and constraints
-- joint-fit weights
-- per-parameter fit controls owned by analysis
-- latest fit-status metadata common to deterministic and Bayesian fits
-- fit-type-specific extensions defined in separate ADRs
-
-Committed parameter values remain in structure or experiment CIF files.
-They are not duplicated into `analysis/analysis.cif`.
-
-### 2. Add a real `_fit_parameter` loop
-
-Introduce a new analysis-owned loop category:
-
-```cif
-loop_
-_fit_parameter.param_unique_name
-_fit_parameter.fit_min
-_fit_parameter.fit_max
-_fit_parameter.fit_bounds_uncertainty_multiplier
-_fit_parameter.start_value
-_fit_parameter.start_uncertainty
-cosio.atom_site.Co1.adp_iso 0.0000 0.1200 4.0 0.0312 0.0021
-cosio.atom_site.Co2.adp_iso 0.0000 0.1200 4.0 0.0312 0.0021
-```
-
-Fields:
-
-- `param_unique_name`
-- `fit_min`
-- `fit_max`
-- `fit_bounds_uncertainty_multiplier`
-- `start_value`
-- `start_uncertainty`
-
-Rationale:
-
-- `fit_min` and `fit_max` are required to restore deterministic and
-  Bayesian fit controls faithfully.
-- `fit_bounds_uncertainty_multiplier` preserves the provenance of
-  uncertainty-derived bounds for restored Bayesian plot annotations.
-- `start_value` and `start_uncertainty` capture the last committed
-  pre-fit scalar state and enable fit recovery workflows, especially in
-  command-line usage.
-- `start_uncertainty` preserves a user-visible pre-fit uncertainty
-  instead of treating it as disposable fit residue.
-
-### 3. Add a generic `_fit_result` single-item category
-
-Persist the latest fit-status metadata shared across fit types in a
-single analysis-owned category:
-
-- `result_kind`
-- `success`
-- `message`
-- `iterations`
-- `fitting_time`
-- `reduced_chi_square`
-
-Suggested CIF fragment:
-
-```cif
-_fit_result.result_kind deterministic
-_fit_result.success yes
-_fit_result.message "Fit converged"
-_fit_result.iterations 37
-_fit_result.fitting_time 1.82
-_fit_result.reduced_chi_square 1.031
-```
-
-`result_kind` distinguishes the latest persisted fit projection, for
-example `deterministic` or `bayesian`.
-
-### 4. Persist only stable fit-status fields here
-
-The `_fit_result` category is for generic status fields that are stable
-across engines and already belong to the result model.
-
-It should not persist backend runtime objects or arbitrary engine
-payloads.
-
-Metrics such as R-factors shown by `display.fit_results()` are derived
-from observed and calculated data and can be recomputed after load when
-needed. They do not need to be part of the first persisted fit-state
-schema.
-
-### 5. Fit-type-specific extensions build on this ADR
-
-This ADR defines the common `analysis.cif` contract for deterministic
-and Bayesian fitting.
-
-Fit-type-specific extensions are layered on top:
-
-- Bayesian persistence extends this with `_bayesian_*` categories and an
-  HDF5 sidecar, as described in `parameter-posterior-summary.md`.
-- Future fit-specific summaries should follow the same pattern: generic
-  shared fields in `_fit_result`, specialized fields in separate
-  categories.
-
-### 6. Restore order is analysis-first, fit-type-second
-
-Load order should be:
-
-1. standard analysis configuration
-2. aliases and constraints
-3. joint-fit weights
-4. `_fit_parameter`
-5. `_fit_result`
-6. fit-type-specific extensions such as `_bayesian_*`
-
-This ensures that generic fit controls are available before restoring
-specialized fit summaries.
-
-### 7. Suggested full `analysis.cif` example
-
-```cif
-_fit.minimizer_type "bumps (dream)"
-_fit.mode single
-
-loop_
-_alias.label
-_alias.param_unique_name
-biso_Co1 cosio.atom_site.Co1.adp_iso
-biso_Co2 cosio.atom_site.Co2.adp_iso
-
-loop_
-_constraint.expression
-"biso_Co2 = biso_Co1"
-
-loop_
-_fit_parameter.param_unique_name
-_fit_parameter.fit_min
-_fit_parameter.fit_max
-_fit_parameter.fit_bounds_uncertainty_multiplier
-_fit_parameter.start_value
-_fit_parameter.start_uncertainty
-cosio.atom_site.Co1.adp_iso 0.0000 0.1200 4.0 0.0312 0.0021
-cosio.atom_site.Co2.adp_iso 0.0000 0.1200 4.0 0.0312 0.0021
-
-_fit_result.result_kind bayesian
-_fit_result.success yes
-_fit_result.message "Sampler converged"
-_fit_result.iterations 3000
-_fit_result.fitting_time 82.4
-_fit_result.reduced_chi_square 1.031
-
-# Bayesian-specific extension categories follow here.
-```
-
-## Consequences
-
-### Positive
-
-- `analysis.cif` becomes the single analysis-owned source of fit state.
-- Deterministic and Bayesian persistence share one common base schema.
-- Fit bounds, bound provenance, and start values survive project
-  reloads.
-- Command-line workflows gain a persisted pre-fit starting state.
-
-### Trade-offs
-
-- The runtime fit-results ADR must be updated because fit state is no
-  longer entirely runtime-only.
-- Analysis persistence becomes more stateful and must be kept in sync
-  with live parameter objects.
-- Some existing serializer assumptions will need refactoring so that
-  `analysis.cif` owns fit metadata rather than individual parameters.
-
-## Deferred Work
-
-- Bayesian-specific categories and HDF5 sidecar details remain in
-  `parameter-posterior-summary.md`.
-- Undo semantics for `start_value` and `start_uncertainty` are defined
-  in a separate ADR.
-- Correlation-matrix persistence is defined in a separate ADR.
diff --git a/docs/dev/adrs/suggestions/fit-output-files-and-data-exports.md b/docs/dev/adrs/suggestions/fit-output-files-and-data-exports.md
new file mode 100644
index 000000000..8705fe355
--- /dev/null
+++ b/docs/dev/adrs/suggestions/fit-output-files-and-data-exports.md
@@ -0,0 +1,172 @@
+# ADR: Fit Output Files and Data Exports
+
+**Status:** Proposed **Date:** 2026-05-18
+
+## Status Note
+
+The current branch already adopts two pieces of this naming scheme:
+
+- sequential deterministic results stay in `analysis/results.csv`
+- Bayesian arrays and plot caches use `analysis/results.h5`
+
+Those decisions now live in
+[Analysis CIF Fit State](../accepted/analysis-cif-fit-state.md). This
+proposal is therefore narrowed to the still-open roles for
+`analysis/data.h5`, `analysis/exports/`, and any extra deterministic
+convenience exports.
+
+## Context
+
+Different fit modes still produce different kinds of reusable output:
+
+- sequential deterministic fits produce a rectangular parameter
+  evolution table, already saved as `analysis/results.csv`
+- Bayesian fits produce posterior samples, diagnostics, predictive
+  arrays, and plot caches, which are too large and structured for CIF or
+  CSV
+- deterministic single and joint fits produce fitted model state,
+  calculated data, reflection tables, residuals, and optional
+  covariance/correlation summaries
+
+The accepted fit-state ADR already standardizes the canonical saved fit
+projection in `analysis/analysis.cif` plus `analysis/results.h5` for
+Bayesian sidecars. What remains open here is whether project save should
+also produce optional archives or user-facing export files beyond that
+accepted baseline.
+
+The project should keep naming consistent and avoid making users extract
+ordinary plotting data from CIF when a clearer CSV export is possible.
+At the same time, CIF remains the canonical model/configuration format,
+and large numerical arrays should not be embedded in
+`analysis/analysis.cif`.
+
+## Decision
+
+### 1. Keep the implemented results baseline
+
+The accepted baseline is:
+
+- `analysis/results.csv` for sequential deterministic fit tables
+- `analysis/results.h5` for large Bayesian arrays and result-derived
+  caches
+
+Any future change to those canonical filenames would need a follow-up
+ADR.
+
+### 2. Reserve separate roles for archives and exports
+
+If extra persisted files are added under `analysis/`, keep their roles
+separate:
+
+- `analysis/data.h5` for optional archived input or measured data.
+- `analysis/exports/` for optional user-facing CSV files intended for
+  external plotting and inspection.
+
+The fit type and saved fit-state manifests stay recorded in
+`analysis/analysis.cif`, principally through `_fit_result.result_kind`
+and the related fit-state categories.
+
+### 3. Sequential deterministic results stay CSV
+
+Sequential deterministic fitting should keep `analysis/results.csv` as
+the canonical table for parameter evolution and extracted metadata.
+
+This file is intentionally CSV because:
+
+- each row naturally corresponds to one sequential fit step
+- users often inspect, filter, and plot it outside EasyDiffraction
+- it should remain easy to diff, copy, and load in spreadsheets
+
+Sequential measured input data may optionally be archived in
+`analysis/data.h5`, but that archive is data, not results. It must not
+replace `analysis/results.csv`.
+
+### 4. Bayesian arrays use `analysis/results.h5`
+
+Single Bayesian fits should store posterior samples, log posterior
+arrays, predictive arrays, and prepared plot caches in
+`analysis/results.h5`.
+
+The previous candidate name `analysis/bayesian_data.h5` remains rejected
+because it mixes fit type with file role and blurs result arrays with
+input data. Bayesian-specific meaning belongs in the CIF manifest and
+HDF5 groups, not the sidecar filename.
+
+### 5. Deterministic single and joint fits may gain CSV exports
+
+For single, joint, and sequential deterministic fits, EasyDiffraction
+should consider optional CSV exports for ordinary plotting data:
+
+- measured data
+- calculated data
+- residuals
+- reflection tables / `refln` categories
+
+These exports are not canonical persistence. They are convenience files
+for users who want to plot or analyze results in external software
+without parsing CIF.
+
+Suggested first layout:
+
+```text
+analysis/
+  analysis.cif
+  results.csv        # sequential deterministic only, when applicable
+  results.h5         # Bayesian and other structured result arrays
+  data.h5            # optional archived measured/input data
+  exports/
+    <experiment>_measured.csv
+    <experiment>_calculated.csv
+    <experiment>_residual.csv
+    <experiment>_reflections.csv
+```
+
+## Fit-Type Mapping
+
+| Fit type                 | Canonical fit state                              | Tabular results              | Large arrays / caches | Optional data archive | Optional exports                |
+| ------------------------ | ------------------------------------------------ | ---------------------------- | --------------------- | --------------------- | ------------------------------- |
+| single deterministic     | `analysis/analysis.cif`                          | open question                | none initially        | none initially        | `analysis/exports/*.csv`        |
+| joint deterministic      | `analysis/analysis.cif`                          | open question                | none initially        | none initially        | `analysis/exports/*.csv`        |
+| sequential deterministic | `analysis/analysis.cif` + `analysis/results.csv` | `analysis/results.csv`       | none initially        | `analysis/data.h5`    | `analysis/exports/*.csv`        |
+| single Bayesian          | `analysis/analysis.cif` + `analysis/results.h5`  | optional summary export only | `analysis/results.h5` | none initially        | optional summary/predictive CSV |
+
+## Open Questions
+
+- Should single and joint deterministic fits write a one-row
+  `analysis/results.csv`, or is their result projection in
+  `analysis/analysis.cif` enough?
+- Should CSV exports be written automatically after fit/save, or only by
+  an explicit export command?
+- What exact CSV column schemas should be used for measured, calculated,
+  residual, and reflection exports?
+- Should exported `refln` CSVs mirror CIF tag names exactly, or use
+  shorter user-facing column names?
+- Should sequential measured data archival in `analysis/data.h5` be
+  opt-in, automatic below a size threshold, or always disabled unless
+  requested?
+- What size threshold and compression policy should control the optional
+  `analysis/data.h5`, and does `analysis/results.h5` need a matching
+  convention?
+- Should external CSV exports be regenerated from canonical CIF/HDF5 on
+  demand rather than stored persistently?
+
+## Consequences
+
+### Positive
+
+- Fit output filenames become role-based and consistent across fit
+  types.
+- Sequential parameter evolution keeps its simple CSV workflow.
+- Bayesian arrays get a generic result sidecar name that can also serve
+  future structured result types.
+- External plotting data can be exposed as plain CSV without weakening
+  CIF as the canonical model format.
+
+### Trade-offs
+
+- The project gains more output-file roles under `analysis/`, so save
+  and cleanup behavior must be explicit.
+- Export CSVs are derived data and must be invalidated or regenerated
+  when the project changes.
+- HDF5 archives introduce size and compression choices that should be
+  resolved before implementation.
diff --git a/docs/dev/adrs/suggestions/parameter-correlation-persistence.md b/docs/dev/adrs/suggestions/parameter-correlation-persistence.md
deleted file mode 100644
index 9c494908b..000000000
--- a/docs/dev/adrs/suggestions/parameter-correlation-persistence.md
+++ /dev/null
@@ -1,114 +0,0 @@
-# ADR: Parameter Correlation Persistence
-
-**Status:** Proposed  
-**Date:** 2026-05-13
-
-## Context
-
-`plot_param_correlations()` can currently visualize either:
-
-- deterministic parameter correlations derived from engine covariance
-- Bayesian correlations derived from posterior samples
-
-After project reload, this correlation information is not available
-unless the underlying runtime objects are rebuilt. For Bayesian fits,
-full posterior samples may not always be restored. For deterministic
-fits, engine covariance is typically not persisted at all.
-
-The correlation matrix is an analysis-owned summary, not model state. It
-therefore belongs in `analysis/analysis.cif`, not in structure or
-experiment CIF files.
-
-## Decision
-
-### 1. Add a `_fit_parameter_correlation` loop category
-
-Persist pairwise parameter correlations in a new analysis-owned loop:
-
-- `source_kind`
-- `param_unique_name_i`
-- `param_unique_name_j`
-- `correlation`
-
-Suggested example:
-
-```cif
-loop_
-_fit_parameter_correlation.source_kind
-_fit_parameter_correlation.param_unique_name_i
-_fit_parameter_correlation.param_unique_name_j
-_fit_parameter_correlation.correlation
-posterior cosio.atom_site.Co1.adp_iso cosio.atom_site.Co2.adp_iso 0.87
-```
-
-`source_kind` records how the correlation was obtained, for example:
-
-- `deterministic`
-- `posterior`
-
-### 2. Store only the upper triangle excluding the diagonal
-
-Each row stores one unordered parameter pair with
-`param_unique_name_i < param_unique_name_j` in a stable ordering.
-
-The diagonal is omitted because it is always 1.0 and can be rebuilt on
-load.
-
-This keeps the CIF loop compact while remaining lossless for the
-correlation matrix.
-
-### 3. Treat the loop as a summary, not a replacement for raw samples
-
-For Bayesian fits, `_fit_parameter_correlation` is a persisted summary.
-It does not replace posterior samples or posterior pair data.
-
-This means:
-
-- correlation heatmaps can be restored from the loop alone
-- posterior pair plots still require posterior samples
-
-### 4. Deterministic and Bayesian fits share the same loop schema
-
-The same loop category is used for both deterministic and Bayesian fit
-results. The distinction is carried by `source_kind`, not by separate
-loop names.
-
-### 5. Suggested restore behavior
-
-On load:
-
-- if `_fit_parameter_correlation` is present, correlation summaries are
-  restored into a lightweight analysis-owned correlation structure
-- if it is absent, correlation plots fall back to whatever live runtime
-  information is available
-
-### 6. Suggested user-facing behavior
-
-```python
-# Restored from analysis.cif when available
-project.display.plotter.plot_param_correlations()
-```
-
-If only the correlation loop is restored, the user still gets the
-correlation heatmap without needing raw posterior samples.
-
-## Consequences
-
-### Positive
-
-- Deterministic and Bayesian correlation summaries survive reload.
-- Correlation heatmaps no longer depend entirely on runtime-only data.
-- The schema is compact and fit-type-agnostic.
-
-### Trade-offs
-
-- The correlation loop is a derived summary, so it must be kept in sync
-  with the latest fit result.
-- Restored correlation data is not enough for posterior pair plots or
-  predictive summaries.
-
-## Deferred Work
-
-- optional storage of covariance matrices in addition to correlations
-- multiple named correlation sources for the same saved project
-- full correlation restoration for pair-plot density surfaces
diff --git a/docs/dev/adrs/suggestions/parameter-posterior-summary.md b/docs/dev/adrs/suggestions/parameter-posterior-summary.md
index 3de39bf37..36c0b979b 100644
--- a/docs/dev/adrs/suggestions/parameter-posterior-summary.md
+++ b/docs/dev/adrs/suggestions/parameter-posterior-summary.md
@@ -1,7 +1,17 @@
-# ADR: Parameter-Level Posterior Projection and Bayesian Persistence
+# ADR: Parameter-Level Posterior Projection
 
-**Status:** Proposed  
-**Date:** 2026-05-13
+**Status:** Proposed **Date:** 2026-05-13
+
+## Status Note
+
+This proposal is narrower than its original draft. The accepted
+[Analysis CIF Fit State](../accepted/analysis-cif-fit-state.md) ADR now
+persists analysis-level posterior summaries in
+`_bayesian_parameter_posterior`, and current Bayesian fits already
+commit the best posterior sample to `parameter.value`. What remains
+undecided is whether parameters should also expose a convenience
+`parameter.posterior` projection and dedicated internal helpers for
+updating fit-derived metadata atomically.
 
 ## Context
 
@@ -13,8 +23,9 @@ Bayesian DREAM currently keeps posterior state only on
 `analysis.fit_results` via `BayesianFitResults`, including
 `posterior_samples`, `posterior_parameter_summaries`,
 `posterior_predictive`, diagnostics, and sampler settings. The accepted
-runtime-fit-results ADR describes this state as runtime-only and not
-serialized.
+runtime-fit-results ADR originally described this state as runtime-only,
+but the accepted fit-state ADR now persists analysis-level posterior
+summaries and cache manifests as a narrower saved projection.
 
 `analysis.fit_results` already changes by analysis type: deterministic
 fits use `FitResults`, while posterior-capable fits such as DREAM use
@@ -59,7 +70,7 @@ Do not add separate flat parameter attributes such as `median`,
 Instead, the parameter-level projection reuses the existing
 `PosteriorParameterSummary` object already produced for
 `BayesianFitResults`. This keeps one grouped summary shape for display,
-inspection, and later persistence.
+inspection, and restore from persisted analysis state.
 
 The summary object currently provides the right level of detail:
 
@@ -192,6 +203,8 @@ while internal fit application installs fresh metadata atomically.
 After a posterior-capable fit, `parameter.value` is committed from the
 best posterior sample.
 
+Current DREAM support already follows this rule.
+
 The best posterior sample is chosen because it is a coherent joint point
 estimate across all free parameters. Marginal medians remain available
 on `parameter.posterior`, but they are summary data rather than the
@@ -211,358 +224,28 @@ Asymmetric interval information is not squeezed into
 `parameter.uncertainty`; it remains available only via
 `parameter.posterior`.
 
-### 10. Persist canonical Bayesian state at analysis level
+### 10. Rebuild posterior from analysis-level state
 
 Canonical Bayesian state is owned by `analysis.fit_results`, not by
-individual parameters.
+individual parameters. The saved fit-state format and restore order are
+defined in `analysis-cif-fit-state.md`.
 
-When the active result is a `BayesianFitResults` instance, persistence
-must save enough data to restore two distinct capability levels:
-
-- summary-only restore for parameter inspection and tables
-- full restore for posterior plots and predictive plots
+The accepted fit-state ADR currently rebuilds analysis-level posterior
+state only. This proposal would add a parameter-level convenience
+projection on top of that restored analysis state.
 
 `parameter.posterior` is never serialized as a per-parameter property.
-It is always rebuilt from the canonical analysis-level persisted data.
-
-### 11. Persist fit-control and Bayesian metadata in `analysis/analysis.cif`
-
-The existing `analysis/analysis.cif` file remains the text metadata
-entry point for analysis persistence.
-
-The persisted fit-control and Bayesian metadata is split into explicit
-CIF categories.
-
-#### 11.1 `_fit_parameter` loop
-
-Stores analysis-owned per-parameter fit metadata that is not currently
-covered by parameter CIF serialization.
-
-This loop exists because the committed parameter CIF representation
-already carries the active `value`, current `free` state, and current
-`uncertainty`, but it does not carry fit bounds, bound provenance, or
-the pre-fit uncertainty snapshot needed by undo. Those fields are
-required for Bayesian plot ranges, pair-plot bound annotations, and
-clean fit rollback after project reload.
-
-Fields:
-
-- `param_unique_name`
-- `fit_min`
-- `fit_max`
-- `fit_bounds_uncertainty_multiplier`
-- `start_value`
-- `start_uncertainty`
-
-`fit_min` and `fit_max` are required for restored Bayesian plotting.
-`fit_bounds_uncertainty_multiplier` is required if restored plots should
-preserve the uncertainty-derived bound annotation exactly. `start_value`
-and `start_uncertainty` are required for clean cross-session undo. If
-omitted, restored fit reports may show `N/A` for `start` and `change`,
-and undo may need to clear uncertainty as a compatibility fallback.
-
-#### 11.2 `_bayesian_result` single item
-
-Stores one saved Bayesian result header with these fields:
-
-- `schema_version`
-- `sampler_name`
-- `point_estimate_name`
-- `success`
-- `sampler_completed`
-- `reduced_chi_square`
-- `fitting_time`
-- `best_log_posterior`
-- `credible_interval_inner`
-- `credible_interval_outer`
-- `has_posterior_samples`
-- `has_posterior_predictive`
-- `sidecar_file`
-
-For the current design, `point_estimate_name` is always `best_sample`.
-
-#### 11.3 `_bayesian_sampler` single item
-
-Stores resolved sampler settings actually used for the run:
-
-- `steps`
-- `burn`
-- `thin`
-- `pop`
-- `parallel`
-- `init`
-- `random_seed`
-
-This persists the existing runtime `sampler_settings` in an explicit,
-schema-driven form rather than as an open-ended key/value mapping.
+It is rebuilt from the analysis-level saved result projection when that
+projection is available.
 
-#### 11.4 `_bayesian_convergence` single item
+Two restore levels matter for the parameter API:
 
-Stores top-level convergence and shape metadata:
+- summary-only restore can populate `parameter.posterior` and fit-result
+  tables
+- full restore can also support posterior plots and predictive plots
 
-- `converged`
-- `max_r_hat`
-- `min_ess_bulk`
-- `n_draws`
-- `n_chains`
-- `n_parameters`
-
-Per-parameter `r_hat` and `ess_bulk` remain in the parameter summary
-loop described below.
-
-#### 11.5 `_bayesian_parameter_posterior` loop
-
-Stores one canonical posterior summary row per sampled parameter. These
-rows are the source used to rebuild `parameter.posterior` on load.
-
-Fields:
-
-- `order_index`
-- `unique_name`
-- `display_name`
-- `best_sample_value`
-- `median`
-- `uncertainty`
-- `interval_68_lower`
-- `interval_68_upper`
-- `interval_95_lower`
-- `interval_95_upper`
-- `ess_bulk`
-- `r_hat`
-
-`order_index` defines the parameter order used by posterior sample
-columns in the sidecar arrays.
-
-#### 11.6 `_bayesian_predictive_dataset` loop
-
-Stores one manifest row per persisted posterior predictive summary.
-
-Fields:
-
-- `experiment_name`
-- `x_axis_name`
-- `x_path`
-- `best_sample_prediction_path`
-- `lower_95_path`
-- `upper_95_path`
-- `lower_68_path`
-- `upper_68_path`
-- `draws_path`
-- `n_x`
-- `n_draws_cached`
-
-This loop tells the loader which arrays to read from the sidecar file
-for each experiment-level predictive summary.
-
-#### 11.7 Suggested CIF fragments
-
-The active parameter value remains in the structure or experiment CIF as
-it does today, for example:
-
-```cif
-_atom_site_U_iso_or_equiv 0.0319(21)
-```
-
-Analysis-owned fit-control and Bayesian metadata then lives in
-`analysis/analysis.cif`, for example:
-
-```cif
-_fit.minimizer_type "bumps (dream)"
-_fit.mode single
-
-loop_
-_fit_parameter.param_unique_name
-_fit_parameter.fit_min
-_fit_parameter.fit_max
-_fit_parameter.fit_bounds_uncertainty_multiplier
-_fit_parameter.start_value
-_fit_parameter.start_uncertainty
-cosio.atom_site.Co1.adp_iso 0.0000 0.1200 4.0 0.0312 0.0021
-cosio.atom_site.Co2.adp_iso 0.0000 0.1200 4.0 0.0312 0.0021
-
-_bayesian_result.schema_version 1
-_bayesian_result.sampler_name dream
-_bayesian_result.point_estimate_name best_sample
-_bayesian_result.success yes
-_bayesian_result.sampler_completed yes
-_bayesian_result.reduced_chi_square 1.031
-_bayesian_result.fitting_time 82.4
-_bayesian_result.best_log_posterior -1542.77
-_bayesian_result.credible_interval_inner 0.68
-_bayesian_result.credible_interval_outer 0.95
-_bayesian_result.has_posterior_samples yes
-_bayesian_result.has_posterior_predictive yes
-_bayesian_result.sidecar_file "bayesian_data.h5"
-
-_bayesian_sampler.steps 3000
-_bayesian_sampler.burn 600
-_bayesian_sampler.thin 1
-_bayesian_sampler.pop 20
-_bayesian_sampler.parallel 0
-_bayesian_sampler.init lhs
-_bayesian_sampler.random_seed 12345
-
-_bayesian_convergence.converged yes
-_bayesian_convergence.max_r_hat 1.01
-_bayesian_convergence.min_ess_bulk 812.4
-_bayesian_convergence.n_draws 2400
-_bayesian_convergence.n_chains 20
-_bayesian_convergence.n_parameters 2
-
-loop_
-_bayesian_parameter_posterior.order_index
-_bayesian_parameter_posterior.unique_name
-_bayesian_parameter_posterior.display_name
-_bayesian_parameter_posterior.best_sample_value
-_bayesian_parameter_posterior.median
-_bayesian_parameter_posterior.uncertainty
-_bayesian_parameter_posterior.interval_68_lower
-_bayesian_parameter_posterior.interval_68_upper
-_bayesian_parameter_posterior.interval_95_lower
-_bayesian_parameter_posterior.interval_95_upper
-_bayesian_parameter_posterior.ess_bulk
-_bayesian_parameter_posterior.r_hat
-0 cosio.atom_site.Co1.adp_iso "Co1 ADP" 0.0319 0.0317 0.0021 0.0298 0.0339 0.0278 0.0361 812.4 1.01
-1 cosio.atom_site.Co2.adp_iso "Co2 ADP" 0.0320 0.0318 0.0020 0.0300 0.0338 0.0281 0.0359 830.7 1.00
-
-loop_
-_bayesian_predictive_dataset.experiment_name
-_bayesian_predictive_dataset.x_axis_name
-_bayesian_predictive_dataset.x_path
-_bayesian_predictive_dataset.best_sample_prediction_path
-_bayesian_predictive_dataset.lower_95_path
-_bayesian_predictive_dataset.upper_95_path
-_bayesian_predictive_dataset.lower_68_path
-_bayesian_predictive_dataset.upper_68_path
-_bayesian_predictive_dataset.draws_path
-_bayesian_predictive_dataset.n_x
-_bayesian_predictive_dataset.n_draws_cached
-hrpt ttheta /predictive/hrpt/x /predictive/hrpt/best_sample_prediction /predictive/hrpt/lower_95 /predictive/hrpt/upper_95 /predictive/hrpt/lower_68 /predictive/hrpt/upper_68 /predictive/hrpt/draws 2500 200
-```
-
-### 12. Persist bulk arrays in `analysis/bayesian_data.h5`
-
-Large numerical arrays are stored outside CIF text in a single sidecar
-file:
-
-- `analysis/bayesian_data.h5`
-
-The sidecar is optional. Summary-only restore remains valid without it.
-
-HDF5 is selected instead of NPZ because the persisted Bayesian payload
-is a structured collection of named datasets with heterogeneous shapes,
-optional groups, and potentially large predictive arrays. HDF5 provides
-explicit hierarchical storage, dataset metadata, selective reads, and a
-better long-term path for compression or chunking. NPZ is simpler, but
-it is flatter and less suitable once the saved Bayesian state grows
-beyond a small set of arrays.
-
-#### 12.1 Required core HDF5 dataset paths when posterior samples are saved
-
-- `posterior_parameter_samples`
-- `posterior_log_posterior`
-- `posterior_draw_index`
-
-Expected array shapes:
-
-- `posterior_parameter_samples`: `(n_draws, n_chains, n_parameters)`
-- `posterior_log_posterior`: `(n_draws, n_chains)` when available
-- `posterior_draw_index`: `(n_draws,)` when available
-
-If `posterior_log_posterior` or `posterior_draw_index` are unavailable,
-their corresponding header flags remain false and the arrays may be
-omitted.
-
-#### 12.2 Predictive dataset keys
-
-Posterior predictive arrays are addressed through the
-`_bayesian_predictive_dataset` manifest rather than inferred from file
-ordering.
-
-Recommended HDF5 dataset naming is:
-
-- `predictive__<experiment>__x`
-- `predictive__<experiment>__best_sample_prediction`
-- `predictive__<experiment>__lower_95`
-- `predictive__<experiment>__upper_95`
-- `predictive__<experiment>__lower_68`
-- `predictive__<experiment>__upper_68`
-- `predictive__<experiment>__draws`
-
-The manifest, not the naming convention, is the source of truth.
-
-Recommended HDF5 group layout is:
-
-- `/posterior/parameter_samples`
-- `/posterior/log_posterior`
-- `/posterior/draw_index`
-- `/predictive/<experiment>/x`
-- `/predictive/<experiment>/best_sample_prediction`
-- `/predictive/<experiment>/lower_95`
-- `/predictive/<experiment>/upper_95`
-- `/predictive/<experiment>/lower_68`
-- `/predictive/<experiment>/upper_68`
-- `/predictive/<experiment>/draws`
-
-The manifest remains the canonical mapping used by the loader, so this
-layout is recommended rather than mandatory.
-
-#### 12.3 What is not persisted in the sidecar
-
-Do not persist backend-specific runtime objects such as `engine_result`,
-the DREAM driver, or ArviZ `InferenceData`.
-
-Those objects can be rebuilt from canonical saved arrays when needed, or
-left unavailable after load.
-
-### 13. Restore flow and partial-availability policy
-
-Persistence must support both full and partial restore.
-
-#### 13.1 Save flow
-
-When `Project.save()` sees `analysis.fit_results` as a
-`BayesianFitResults` instance:
-
-1. It writes standard analysis configuration to `analysis/analysis.cif`.
-2. It appends `_fit_parameter`, `_bayesian_result`, `_bayesian_sampler`,
-   `_bayesian_convergence`, and `_bayesian_parameter_posterior`.
-3. If posterior predictive summaries are available, it also writes the
-   `_bayesian_predictive_dataset` manifest.
-4. If posterior sample arrays or predictive arrays are available, it
-   writes `analysis/bayesian_data.h5`.
-
-#### 13.2 Load flow
-
-When `Project.load()` restores `analysis/analysis.cif`:
-
-1. Standard analysis configuration is restored first.
-2. If `_fit_parameter` is present, fit bounds, bound provenance, and
-   optional pre-fit scalar snapshots are restored by matching
-   `param_unique_name` to live parameters.
-3. If Bayesian categories are present, a lightweight
-   `BayesianFitResults` instance is rebuilt from persisted metadata and
-   parameter summary rows.
-4. `parameter.posterior` is rebuilt by matching summary rows to live
-   parameters via `unique_name`.
-5. `parameter.value` and `parameter.uncertainty` continue to come from
-   the normal project serialization path; Bayesian restore does not
-   overwrite them.
-6. If `analysis/bayesian_data.h5` exists and matches the manifest,
-   `posterior_samples` and `posterior_predictive` are restored.
-7. If the sidecar is missing or incomplete, the restore degrades to
-   summary-only mode without failing the whole project load.
-
-#### 13.3 Partial restore behavior
-
-The chosen partial-restore policy is:
-
-- `parameter.posterior` and `display.fit_results()` remain available
-  from saved metadata and summaries.
-- Bayesian-only plots requiring canonical posterior arrays remain
-  unavailable unless those arrays were restored successfully.
-- Missing sidecar data should produce a clear warning, not a hard load
-  failure.
+If the saved project has no analysis-level posterior summary for a
+parameter, `parameter.posterior` remains `None`.
 
 Example user access after load:
 
@@ -574,28 +257,18 @@ posterior = param.posterior
 
 if posterior is not None:
   print(posterior.best_sample_value)
-    print(posterior.uncertainty)
-    print(posterior.interval_68)
+  print(posterior.uncertainty)
+  print(posterior.interval_68)
 
 project.analysis.display.fit_results()
 ```
 
-Example plotting behavior after load:
-
-```python
-# Works with summary-only restore
-project.analysis.display.fit_results()
+Posterior plot availability after load follows the fit-state restore
+level defined in `analysis-cif-fit-state.md`.
 
-# Requires canonical posterior arrays from analysis/bayesian_data.h5
-project.display.plotter.plot_posterior_pairs()
-project.display.plotter.plot_param_distribution(param)
-project.display.plotter.plot_posterior_predictive(expt_name='hrpt')
-```
+### 11. Keep parameter posterior data rebuilt, not duplicated
 
-### 14. Keep parameter posterior data rebuilt, not duplicated
-
-`parameter.posterior` is always rebuilt from the canonical
-`_bayesian_parameter_posterior` loop in `analysis/analysis.cif`.
+`parameter.posterior` is always rebuilt from analysis-level fit state.
 
 Do not serialize posterior summaries again inside each parameter's own
 CIF representation. Duplicating the same posterior summary data across
@@ -610,15 +283,10 @@ structure and experiment files would create multiple sources of truth.
   into one optional object rather than many flat attributes.
 - `uncertainty` and posterior metadata become clearly fit-owned rather
   than mixed user-editable and fit-editable state.
-- `analysis.fit_results` remains the canonical source for full Bayesian
-  state.
-- `analysis/analysis.cif` becomes the home for fit-control metadata that
-  does not belong in structure or experiment CIF files.
-- Bayesian save/load gains a clear split between text metadata in
-  `analysis/analysis.cif` and bulk numerical arrays in
-  `analysis/bayesian_data.h5`.
-- Partial restore works for summaries even when full posterior arrays
-  are absent.
+- `analysis.fit_results` remains the canonical runtime source for full
+  Bayesian state.
+- Partial restore can still expose parameter summaries when
+  analysis-level saved state contains them.
 - The design matches the current rule that `value` is the only active
   scalar used for calculations.
 
@@ -632,8 +300,6 @@ structure and experiment files would create multiple sources of truth.
   application code must be updated to use dedicated internal helpers
   rather than a mix of `_set_value_from_minimizer(...)` and public
   uncertainty assignment.
-- Bayesian persistence now spans both CIF metadata and a binary sidecar,
-  so save/load code must validate consistency between the two.
 
 ## Layering and Ownership
 
@@ -647,14 +313,8 @@ type-only import is acceptable.
 
 ## Deferred Work
 
-This ADR now defines persistence for one canonical saved Bayesian run.
-
-It still defers:
+This ADR defines the parameter-level posterior projection. It defers:
 
-- support for multiple saved Bayesian runs per project
-- plot-ready cache layers beyond canonical posterior and predictive data
-- persistence-time compression or chunking strategies beyond the first
-  HDF5 sidecar implementation
 - persistence for future posterior-capable minimizers beyond DREAM
 - enabling currently unsupported single-crystal predictive draw plots
 
@@ -665,18 +325,13 @@ It still defers:
 - Existing `PosteriorParameterSummary` instances should be reused rather
   than copied into a second summary type unless implementation reveals a
   concrete layering problem that cannot be resolved cleanly.
-- `parameter.posterior` should always be rebuilt from restored canonical
+- `parameter.posterior` should always be rebuilt from analysis-level
   Bayesian data rather than serialized redundantly at parameter level.
-- The sidecar reader and writer should be isolated behind explicit
-  serializer helpers instead of being implemented inline in
-  `Project.save()` and `Project.load()`.
 
 ## Chosen Defaults
 
 - `parameter.value` remains committed to the best posterior sample after
   posterior fits.
-- If a project is loaded without full posterior arrays, restoring only
+- If a project is loaded with only posterior summaries, restoring
   `parameter.posterior` is acceptable for table display and parameter
   inspection.
-- Posterior plotting remains unavailable unless the canonical Bayesian
-  containers needed by those plots are also restored.
diff --git a/docs/dev/adrs/suggestions/python-cif-category-correspondence.md b/docs/dev/adrs/suggestions/python-cif-category-correspondence.md
index 6f970a0e2..191f18e59 100644
--- a/docs/dev/adrs/suggestions/python-cif-category-correspondence.md
+++ b/docs/dev/adrs/suggestions/python-cif-category-correspondence.md
@@ -24,13 +24,20 @@ one-to-one correspondence for project-owned singleton categories:
 ```text
 project.info.title        -> project.cif: _info.title
 project.rendering.engine  -> project.cif: _rendering.engine
-project.verbosity.level   -> project.cif: _verbosity.level
+project.verbosity.fit     -> project.cif: _verbosity.fit
 ```
 
 The design question is whether this rule should be applied only to
 project-level configuration, or more broadly across analysis,
 experiments, structures, and calculated data.
 
+The accepted project-facade decision keeps `Project` as the public root
+and keeps `project.cif` as the singleton project configuration file. It
+also keeps `_project.*` as the semantic CIF category for scientific
+project information and rejects `_meta.*` for that purpose. This ADR
+therefore must not reintroduce the rejected `Workspace` rename,
+`workspace.cif`, or `_meta.project_*` tags as incidental cleanup.
+
 ## Scope Of Comparison
 
 The comparison below is category-level and public-API oriented. It lists
@@ -48,7 +55,7 @@ to objects reached from the current `Project` root, for example
 | Current Python surface              | Current saved location   | Current CIF block form | Notes                                                                               |
 | ----------------------------------- | ------------------------ | ---------------------- | ----------------------------------------------------------------------------------- |
 | `project.info`, `project.rendering` | `project.cif`            | bare categories        | Project-level singleton config.                                                     |
-| `project.verbosity`                 | not persisted            | none                   | Runtime-only string property backed by `VerbosityEnum`; no `_verbosity` category.   |
+| `project.verbosity`                 | `project.cif`            | bare category          | Project-owned fit-output verbosity category backed by `VerbosityEnum`.              |
 | `project.structures[name]`          | `structures/<name>.cif`  | `data_<name>`          | Each structure is one CIF data block.                                               |
 | `project.experiments[name]`         | `experiments/<name>.cif` | `data_<name>`          | Each experiment is one CIF data block.                                              |
 | `project.analysis`                  | `analysis/analysis.cif`  | bare categories        | Loader also accepts legacy root-level `analysis.cif`.                               |
@@ -58,17 +65,17 @@ to objects reached from the current `Project` root, for example
 
 ### Project-Level Configuration
 
-| Current Python path              | Current CIF path          | Match? | Notes                                                                                                                           |
-| -------------------------------- | ------------------------- | ------ | ------------------------------------------------------------------------------------------------------------------------------- |
-| `project.info.name`              | `_project.id`             | No     | Python uses user-facing `name`; CIF uses `id`; category is `info` in Python but `_project` in CIF.                              |
-| `project.info.title`             | `_project.title`          | Partly | Field name matches, category name does not.                                                                                     |
-| `project.info.description`       | `_project.description`    | Partly | Field name matches, category name does not.                                                                                     |
-| `project.info.created`           | `_project.created`        | Partly | Field name matches, category name does not.                                                                                     |
-| `project.info.last_modified`     | `_project.last_modified`  | Partly | Field name matches, category name does not.                                                                                     |
-| `project.info.path`              | none                      | No     | Runtime storage path, not a CIF field.                                                                                          |
-| `project.rendering.chart_engine` | `_rendering.chart_engine` | Yes    | Direct category and field mapping.                                                                                              |
-| `project.rendering.table_engine` | `_rendering.table_engine` | Yes    | Direct category and field mapping.                                                                                              |
-| `project.verbosity`              | none                      | No     | Runtime-only string convenience property; current code has no `project.verbosity.level` category and no `_verbosity.level` tag. |
+| Current Python path              | Current CIF path          | Match? | Notes                                                                                              |
+| -------------------------------- | ------------------------- | ------ | -------------------------------------------------------------------------------------------------- |
+| `project.info.name`              | `_project.id`             | No     | Python uses user-facing `name`; CIF uses `id`; category is `info` in Python but `_project` in CIF. |
+| `project.info.title`             | `_project.title`          | Partly | Field name matches, category name does not.                                                        |
+| `project.info.description`       | `_project.description`    | Partly | Field name matches, category name does not.                                                        |
+| `project.info.created`           | `_project.created`        | Partly | Field name matches, category name does not.                                                        |
+| `project.info.last_modified`     | `_project.last_modified`  | Partly | Field name matches, category name does not.                                                        |
+| `project.info.path`              | none                      | No     | Runtime storage path, not a CIF field.                                                             |
+| `project.rendering.chart_engine` | `_rendering.chart_engine` | Yes    | Direct category and field mapping.                                                                 |
+| `project.rendering.table_engine` | `_rendering.table_engine` | Yes    | Direct category and field mapping.                                                                 |
+| `project.verbosity.fit`          | `_verbosity.fit`          | Yes    | Direct category and field mapping for fitting process output verbosity.                            |
 
 ### Analysis Configuration
 
@@ -218,18 +225,30 @@ This ADR does not propose renaming the public root object. The current
 root object is already `Project`; the proposal is about category and tag
 correspondence inside project-owned singleton configuration.
 
-Target project-level mappings if the current Python names are kept:
+The accepted baseline is:
+
+```text
+project.info.<field> -> project.cif: _project.<field>
+```
 
-| Python path                      | Target CIF path           | Current state                              |
-| -------------------------------- | ------------------------- | ------------------------------------------ |
-| `project.info.name`              | `_info.name`              | Currently `_project.id`.                   |
-| `project.info.title`             | `_info.title`             | Currently `_project.title`.                |
-| `project.info.description`       | `_info.description`       | Currently `_project.description`.          |
-| `project.info.created`           | `_info.created`           | Currently `_project.created`.              |
-| `project.info.last_modified`     | `_info.last_modified`     | Currently `_project.last_modified`.        |
-| `project.rendering.chart_engine` | `_rendering.chart_engine` | Already matches.                           |
-| `project.rendering.table_engine` | `_rendering.table_engine` | Already matches.                           |
-| `project.verbosity.level`        | `_verbosity.level`        | Currently no persisted verbosity category. |
+Future one-to-one correspondence work may still discuss whether the
+public identity field should be `name` or `id`, whether verbosity should
+gain additional coverage-specific fields, and whether rendering should
+keep separate chart and table engine fields.
+
+Possible strict-correspondence target if a future ADR explicitly changes
+the accepted `_project.*` baseline:
+
+| Python path                      | Target CIF path           | Current state                                    |
+| -------------------------------- | ------------------------- | ------------------------------------------------ |
+| `project.info.name`              | `_info.name`              | Currently `_project.id`.                         |
+| `project.info.title`             | `_info.title`             | Currently `_project.title`.                      |
+| `project.info.description`       | `_info.description`       | Currently `_project.description`.                |
+| `project.info.created`           | `_info.created`           | Currently `_project.created`.                    |
+| `project.info.last_modified`     | `_info.last_modified`     | Currently `_project.last_modified`.              |
+| `project.rendering.chart_engine` | `_rendering.chart_engine` | Already matches.                                 |
+| `project.rendering.table_engine` | `_rendering.table_engine` | Already matches.                                 |
+| `project.verbosity.fit`          | `_verbosity.fit`          | Implemented direct fit-output verbosity mapping. |
 
 Alternative target if the project identity field should be called `id`
 rather than `name`:
@@ -253,15 +272,24 @@ repository can optimize them for API/persistence symmetry.
 ### `project.cif` Scopes Generic Categories
 
 `_info.title` is generic in isolation, but inside `project.cif` it reads
-as project information. This is similar to `_verbosity.level`: the file
-scope tells the reader this is project-level verbosity.
+as project information. This is similar to `_verbosity.fit`: the file
+scope tells the reader this is project-level verbosity, and the field
+name identifies the fitting-process coverage.
 
 ### The Current `Project` Root Already Matches User Language
 
 The current public root object is already `Project`. Keeping it avoids a
-user-facing `workspace.project.*` nesting and aligns with scientific
-workflows where a project is the container for structures, experiments,
-analysis, and saved files.
+broad user-facing root rename and aligns with scientific workflows where
+a project is the container for structures, experiments, analysis, and
+saved files.
+
+### `_project.*` Is More Semantic Than `_meta.*`
+
+The project-information category stores the scientific project identity,
+title, description, and timestamps. `_project.id` and `_project.title`
+say that directly, while `_meta.project_id` and `_meta.project_title`
+make the CIF less domain-oriented and repeat the concept in every item
+name.
 
 ### Scientific CIF/Domain Categories Should Stay Domain-Oriented
 
@@ -291,9 +319,9 @@ unless a separate ADR changes the underlying API pattern.
 - `_info.*` is less self-describing if copied out of `project.cif`.
 - Existing `_project.*` project files would need migration or a
   deliberate compatibility decision.
-- If verbosity is persisted, `project.verbosity` would either need to
-  become a category object or remain as a convenience alias for a new
-  `project.verbosity.level` field.
+- Persisted verbosity is now a category object. The initial field is
+  `project.verbosity.fit`, leaving room for future coverage-specific
+  verbosity fields.
 - Collapsing rendering to `project.rendering.engine` would simplify the
   API, but only if chart and table renderers are intended to share one
   backend choice.
@@ -302,13 +330,9 @@ unless a separate ADR changes the underlying API pattern.
 
 - Should the project identity remain `project.info.name`, or should it
   become `project.info.id` to mirror the saved identifier field?
-- Should project metadata move from `_project.*` to `_info.*`, or is
-  `_project.*` clearer even inside `project.cif`?
 - Should `project.rendering.chart_engine` and
   `project.rendering.table_engine` remain separate, or should the public
   API and CIF collapse to one `engine` field?
-- Should `_project.*` be accepted as a read-only legacy fallback when
-  loading older saved projects?
 - Should `project.verbosity = 'short'` remain as a convenience alias for
-  `project.verbosity.level = 'short'`, or should strict correspondence
+  `project.verbosity.fit = 'short'`, or should strict correspondence
   remove the alias?
diff --git a/docs/dev/adrs/suggestions/undo-fit.md b/docs/dev/adrs/suggestions/undo-fit.md
index 578d43a38..cd658aba9 100644
--- a/docs/dev/adrs/suggestions/undo-fit.md
+++ b/docs/dev/adrs/suggestions/undo-fit.md
@@ -1,57 +1,61 @@
 # ADR: Undo Fit
 
 **Status:** Proposed  
-**Date:** 2026-05-13
+**Date:** 2026-05-18
+
+## Status Note
+
+The rollback anchors described here are already persisted and restored.
+Current code saves `_fit_parameter.start_value` and
+`_fit_parameter.start_uncertainty` in `analysis/analysis.cif`, and the
+CLI already reserves `PROJECT_DIR undo`, but no rollback operation is
+implemented yet.
 
 ## Context
 
-The new `_fit_parameter.start_value` and
-`_fit_parameter.start_uncertainty` fields in `analysis/analysis.cif`
-capture the last committed pre-fit scalar state for each fitted
-parameter. This is useful when a minimization run produces a poor result
-and the user wants to return to the state from immediately before the
-fit.
+The accepted fit-state persistence design now stores
+`_fit_parameter.start_value` and `_fit_parameter.start_uncertainty` in
+`analysis/analysis.cif`. Those fields capture the last committed pre-fit
+scalar state for each fitted parameter and are the essential rollback
+anchors for any undo feature.
 
-This need is especially important in command-line workflows, where the
-user may save a project after a bad fit and reopen it later expecting a
-simple way to roll back to the pre-fit state.
+This branch also introduced project-first CLI routing and reserved a
+top-level `undo` command shape, but the command is still only a
+placeholder. The actual rollback semantics are still undecided.
 
-However, these snapshots alone do not define undo semantics. The API
-owner, rollback scope, and interaction with fit-derived metadata must be
-explicit.
+Parameter-level posterior access remains a separate proposal. Undo must
+not depend on `parameter.posterior` existing.
 
 ## Decision
 
 ### 1. Add an analysis-owned `undo_fit()` operation
 
-The rollback operation belongs on `Analysis`, for example:
+The rollback operation belongs on `Analysis`:
 
 ```python
 project.analysis.undo_fit()
 ```
 
-`Analysis` owns the fit lifecycle, fit metadata, and persisted
-`analysis.cif` state, so it is the correct owner for this operation.
+`Analysis` owns fit execution, fit metadata, and the persisted fit-state
+projection, so it is the correct public owner.
 
-### 2. Initial undo scope is scalar rollback plus posterior clear
+### 2. Initial undo scope is scalar rollback plus fit-state clear
 
 The first undo implementation restores each fitted parameter's saved
-pre-fit scalar state and clears fit state that belongs only to the
-discarded fit.
-
-It does not attempt to restore every possible runtime detail of the
-previous fit result.
+pre-fit scalar state and clears fit-derived state that belongs only to
+the discarded fit.
 
 After `undo_fit()`:
 
 - `parameter.value` is restored from `_fit_parameter.start_value`
 - `parameter.uncertainty` is restored from
   `_fit_parameter.start_uncertainty`
-- `parameter.posterior` is cleared
 - `analysis.fit_results` is cleared
+- persisted fit-state summaries and Bayesian caches for the discarded
+  fit are cleared
 
-This gives the user a safe, predictable return to the pre-fit visible
-parameter state without pretending to restore a full historical result.
+If a future `parameter.posterior` API exists, undo should clear that
+projection too. It is not a prerequisite for the initial implementation.
 
 If an older saved project lacks `start_uncertainty`, clearing
 `parameter.uncertainty` remains an acceptable compatibility fallback.
@@ -67,51 +71,20 @@ The initial undo operation does not revert:
 - fit mode
 - joint-fit weights
 
-These are analysis configuration, not fit output.
+These belong to analysis configuration, not fit output.
 
 ### 4. Undo is single-level for now
 
-Only the latest saved pre-fit state is addressable.
-
-The initial API does not create a stack of historical fits. Supporting
-multiple undo levels would require a dedicated snapshot history design
-and is deferred.
-
-### 5. Persisted scalar snapshots are the rollback anchors
-
-The minimum persisted state required for clean cross-session undo is the
-pair of `_fit_parameter.start_value` and
-`_fit_parameter.start_uncertainty` defined in
-`analysis-cif-fit-state.md`.
-
-If a parameter has no saved `start_value`, `undo_fit()` leaves that
-parameter unchanged.
-
-If a parameter has no saved `start_uncertainty`, `undo_fit()` may clear
-that parameter's uncertainty as a compatibility fallback for older saved
-projects.
-
-### 6. Suggested user flow
-
-```python
-project.analysis.fit()
-
-# Decide that the latest fit should be discarded.
-project.analysis.undo_fit()
-
-# Save the recovered state if desired.
-project.save()
-```
+Only the latest saved pre-fit snapshot is addressable. Multi-level undo
+and redo require a dedicated snapshot-history design and remain
+deferred.
 
-### 7. Add a top-level `undo-fit` CLI command
+### 5. CLI exposure follows the project-first command style
 
-Because command-line recovery is one of the main motivations for this
-feature, undo must also be exposed through the existing top-level CLI.
-
-Suggested command:
+The command-line surface should follow the current CLI style:
 
 ```bash
-python -m easydiffraction undo-fit PROJECT_DIR
+python -m easydiffraction PROJECT_DIR undo
 ```
 
 This command should:
@@ -119,44 +92,35 @@ This command should:
 - load the saved project from `PROJECT_DIR`
 - execute `project.analysis.undo_fit()`
 - save the recovered state back to the same project directory by default
-- support `--dry` to perform the rollback in memory without overwriting
-  project files
-- emit a clear message describing whether the latest fit snapshot was
-  successfully discarded
-
-Suggested dry-run form:
-
-```bash
-python -m easydiffraction undo-fit PROJECT_DIR --dry
-```
+- support `--dry` to preview the rollback without overwriting files
+- fail with a clear non-zero exit status when no usable undo snapshot is
+  available
 
-If the project does not contain a usable undo snapshot, the command
-should fail with a clear non-zero exit status instead of silently doing
-nothing.
+Compatibility aliases may remain if the CLI supports them, but the
+project-first form is the canonical user-facing syntax.
 
 ## Consequences
 
 ### Positive
 
-- Users gain a simple recovery path after a poor fit.
-- The feature works naturally with saved projects and both Python and
-  command-line workflows.
-- The initial scope stays small and does not require full historical fit
-  snapshots.
+- The accepted fit-state persistence already provides the minimum saved
+  anchors required for cross-session undo.
+- Users gain a predictable recovery path after a poor fit without
+  needing full historical fit snapshots.
+- The feature aligns naturally with saved-project workflows in both
+  Python and the CLI.
 
 ### Trade-offs
 
-- Undo restores pre-fit scalar parameter state, not a full historical
-  `fit_results` object.
-- Older saved projects that do not carry `start_uncertainty` may still
-  fall back to clearing uncertainty.
+- Undo restores visible scalar parameter state, not a full historical
+  runtime result object.
+- Older saved projects may still need the uncertainty-clearing fallback.
 - Multi-level undo remains unsupported.
 
 ## Deferred Work
 
-- exact restoration of previous posterior-derived projections beyond the
-  scalar parameter snapshot
-- multi-level undo / redo
-- user-facing confirmation or preview APIs
-- rollback of fit-type-specific persisted summaries beyond parameter
-  values
+- exact restoration of previous posterior-derived displays beyond the
+  scalar rollback anchors
+- multi-level undo and redo
+- confirmation or preview UX beyond `--dry`
+- any dependency on a future `parameter.posterior` API
diff --git a/docs/dev/adrs/suggestions/workspace-root-project-category.md b/docs/dev/adrs/suggestions/workspace-root-project-category.md
deleted file mode 100644
index a6c94cba6..000000000
--- a/docs/dev/adrs/suggestions/workspace-root-project-category.md
+++ /dev/null
@@ -1,380 +0,0 @@
-# ADR: Workspace Root and Project Information Category
-
-**Status:** Proposed  
-**Date:** 2026-05-17
-
-## Context
-
-The current public root object is `Project`. It acts as the top-level
-facade for an EasyDiffraction working session:
-
-```python
-project = ed.Project(name='lbco_hrpt')
-project.structures
-project.experiments
-project.analysis
-project.display
-project.summary
-project.save()
-```
-
-The same word, "project", is also the natural CIF category name for
-information about the scientific project:
-
-```cif
-_project.id
-_project.title
-_project.description
-_project.created
-_project.last_modified
-```
-
-This creates a naming conflict. The root object is a broad runtime
-facade, while the `_project.*` category is only information about the
-scientific project. Using the same name for both makes category naming
-awkward:
-
-```python
-project.project_info.title
-project.config.project_info.title
-project.project.title
-```
-
-At the same time, replacing `_project.*` with a generic `_meta.*`
-category would weaken the CIF model:
-
-```cif
-_meta.project_id
-_meta.project_title
-```
-
-The category name `meta` is too generic. It forces each item name to
-repeat what the category should already communicate. The existing
-`_project.id` and `_project.title` names are more semantic and better
-aligned with the repository rule to follow CIF naming unless there is a
-clearly better API.
-
-The design question is therefore:
-
-- should the top-level runtime object remain `Project`, and project
-  information move to a different category such as `meta`;
-- or should the top-level runtime object be renamed to `Workspace`, so
-  `project` can be used cleanly for project information?
-
-## Decision
-
-Rename the top-level runtime facade from `Project` to `Workspace`.
-
-Use `project` as the public project-information category under the
-workspace:
-
-```python
-workspace = ed.Workspace(project_id='lbco_hrpt')
-workspace.project.id
-workspace.project.title = 'La0.5Ba0.5CoO3 at HRPT@PSI'
-workspace.rendering.table_engine = 'rich'
-workspace.verbosity = 'short'
-workspace.structures
-workspace.experiments
-workspace.analysis
-```
-
-Persist workspace-level singleton categories in `workspace.cif`:
-
-```cif
-_project.id
-_project.title
-_project.description
-_project.created
-_project.last_modified
-
-_rendering.chart_engine
-_rendering.table_engine
-
-_verbosity.level
-```
-
-The saved directory is a workspace directory whose filesystem name is
-chosen by the user. The canonical layout is:
-
-```text
-<workspace-dir>/
-|-- workspace.cif
-|-- structures/
-|   `-- cosio.cif
-|-- experiments/
-|   `-- d20.cif
-|-- analysis/
-|   `-- analysis.cif
-`-- summary/
-    `-- summary.cif
-```
-
-Do not introduce `_meta.*` CIF tags. Do not use `project.cif`,
-`config.cif`, or `meta.cif` as the primary singleton configuration file
-in the target layout.
-
-The intended naming split is:
-
-```text
-Workspace
-|-- project      # information about the scientific project
-|-- rendering    # rendering preferences
-|-- verbosity    # console/output verbosity preference
-|-- structures   # real structure datablocks
-|-- experiments  # real experiment datablocks
-|-- analysis     # analysis section
-|-- display      # display facade
-`-- summary      # summary/report facade
-```
-
-## Rationale
-
-### `Workspace` better describes the top-level facade
-
-The top-level object is more than project metadata. It owns active
-collections, analysis state, display helpers, save/load behavior, and
-runtime orchestration. `Workspace` describes that broader role without
-consuming the domain word `project`.
-
-The name is also familiar in scientific software. It commonly means an
-active analysis environment, a data/model container, or a working area.
-That is close to the role of the current EasyDiffraction root object.
-
-### `project` is the right category name for project information
-
-Project information is not generic metadata. It is specifically the
-identity, title, description, and timestamps of the scientific project.
-
-This reads cleanly:
-
-```python
-workspace.project.title
-```
-
-and maps directly to clean CIF:
-
-```cif
-_project.title
-```
-
-### `_meta.project_title` is weaker than `_project.title`
-
-The `_meta` category would make the CIF less domain-oriented. It also
-creates longer and more repetitive item names:
-
-```cif
-_meta.project_id
-_meta.project_title
-_meta.project_description
-```
-
-The existing `_project.*` tags are clearer:
-
-```cif
-_project.id
-_project.title
-_project.description
-```
-
-### This keeps layer-specific consistency
-
-After this decision, each layer has a clear rule:
-
-| Layer           | Rule                           | Example             |
-| --------------- | ------------------------------ | ------------------- |
-| Runtime root    | working-session facade         | `Workspace`         |
-| Public category | semantic category name         | `workspace.project` |
-| CIF category    | semantic CIF category          | `_project.*`        |
-| Config file     | workspace singleton categories | `workspace.cif`     |
-
-This avoids one-off aliases such as `project.info` while preserving
-semantic CIF names.
-
-### `workspace.cif` is clearer than `project.cif`, `config.cif`, or `meta.cif`
-
-The file stores singleton settings owned by the workspace: scientific
-project information, rendering preferences, and verbosity. `project.cif`
-overloads the project name again, while `config.cif` and `meta.cif` are
-generic. `workspace.cif` names the owning layer and lets each category
-inside the file keep its domain-specific name.
-
-## Consequences
-
-### Positive
-
-- The root object and project-information category no longer share the
-  same conceptual name.
-- Public access becomes uniform: `workspace.project`,
-  `workspace.rendering`, `workspace.verbosity`, `workspace.analysis`.
-- CIF stays semantic and does not introduce `_meta.*`.
-- Workspace-level preferences such as rendering and verbosity are saved
-  with the workspace instead of being hidden runtime-only state.
-- Project information can use short item names such as `id`, `title`,
-  and `description`.
-- The top-level facade name better reflects active runtime
-  orchestration.
-
-### Negative
-
-- This is a breaking public API change.
-- Tutorials, scripts, tests, docs, type hints, and imports must be
-  updated from `Project` to `Workspace`.
-- Existing saved directories using `project.cif` must be migrated to
-  `workspace.cif` if no compatibility loader is kept.
-- Existing code that expected verbosity to be runtime-only must account
-  for it round-tripping through `workspace.cif`.
-- Users familiar with `Project` must learn the new root name.
-- `Workspace` can be confused with a filesystem workspace in some
-  ecosystems, so documentation must define it clearly as the active
-  EasyDiffraction working object.
-
-## Compatibility Policy
-
-EasyDiffraction is in beta, and repository instructions say not to keep
-legacy shims by default.
-
-Therefore the target implementation should not add a permanent
-`Project = Workspace` alias unless explicitly approved before
-implementation.
-
-The migration plan still includes a review gate before removing the old
-`Project` public symbol, because this is a user-facing breaking change.
-
-## Alternatives Considered
-
-### Keep `Project` root and rename project information to `meta`
-
-Rejected.
-
-Example:
-
-```python
-project.meta.project_title
-```
-
-```cif
-_meta.project_title
-```
-
-This keeps the root class stable, but it weakens the category model.
-`meta` is too broad, and the CIF item names become repetitive.
-
-### Keep `Project` root and use `project.config.project`
-
-Rejected for now.
-
-Example:
-
-```python
-project.config.project.title
-```
-
-This is technically consistent, but it still repeats `project` at
-different semantic layers. It also adds depth to common user workflows.
-It is a reasonable fallback if the public root rename is rejected.
-
-### Keep `Project` root and use `project.info`
-
-Rejected for the target design.
-
-Example:
-
-```python
-project.info.title
-```
-
-This is readable, but it preserves a special-case category alias. The
-current goal is stronger consistency between public categories and CIF
-category concepts.
-
-### Use `project.cif`, `config.cif`, or `meta.cif` for singleton settings
-
-Rejected for the target layout.
-
-`project.cif` repeats the overloaded term that this migration removes.
-`config.cif` and `meta.cif` are too generic and do not say which layer
-owns the settings. `workspace.cif` is more explicit while still allowing
-semantic categories such as `_project`, `_rendering`, and `_verbosity`
-inside the file.
-
-### Rename only internal files and keep public API unchanged
-
-Rejected for the target design.
-
-This improves implementation clarity but does not solve the public
-naming inconsistency.
-
-### Use `Study` instead of `Workspace`
-
-Rejected.
-
-`Study` is a plausible scientific term, but it is less established for
-an active computational container. It also does not map as naturally to
-save/load, display, and analysis orchestration.
-
-## Implementation Notes
-
-The implementation should follow:
-
-```text
-docs/dev/plans/workspace-root-project-category.md
-```
-
-The high-level migration is:
-
-1. Rename the root facade `Project` to `Workspace`.
-2. Rename the project package/module surface from `project` to
-   `workspace`.
-3. Rename `ProjectConfig` to `WorkspaceConfig`.
-4. Rename project-level category access from `info` to `project`.
-5. Rename project-information `name` access to `id`, matching
-   `_project.id`.
-6. Move the storage path to `Workspace.path`, because it describes the
-   saved workspace directory rather than project information.
-7. Keep the information category class named `ProjectInfo` unless a
-   later decision chooses `ProjectMetadata`.
-8. Keep CIF tags `_project.*` and `_rendering.*`.
-9. Rename saved singleton config file from `project.cif` to
-   `workspace.cif`.
-10. Persist workspace verbosity in `workspace.cif` as
-    `_verbosity.level`, owned by a first-class `Verbosity` category
-    under `WorkspaceConfig` (parallel to `Rendering`).
-11. Update code, tests, scripts, tutorials, docs, and ADR references.
-
-## Post-Implementation ADR Update
-
-This ADR must be updated after the migration plan is implemented.
-
-When implementation is complete:
-
-1. Change status from `Proposed` to `Accepted and implemented`.
-2. Record the final public API and saved file layout.
-3. Record whether a temporary or permanent `Project` compatibility alias
-   was approved.
-4. Record any deviations from the migration plan.
-5. Move this file from `docs/dev/adrs/suggestions/` to
-   `docs/dev/adrs/accepted/` if the decision is accepted.
-6. Update `docs/dev/adrs/index.md` and related accepted ADRs if the ADR
-   map changes.
-7. Update or close related items in `docs/dev/issues/open.md`.
-
-## Acceptance Criteria
-
-This ADR is satisfied when:
-
-- `ed.Workspace` is the public top-level facade.
-- `ed.Project` is removed unless explicitly approved as an alias.
-- the public project-information category is `workspace.project`.
-- project identity is exposed as `workspace.project.id`.
-- the saved directory path is exposed as `workspace.path`.
-- the public rendering category is `workspace.rendering`.
-- saved singleton configuration lives in `workspace.cif`.
-- `workspace.cif` uses `_project.*`, `_rendering.*`, and
-  `_verbosity.level` tags.
-- workspace verbosity is owned by a registered `Verbosity` category
-  alongside `Rendering`.
-- `ProjectInfo.path` is removed; the saved directory path is exposed
-  only as `workspace.path`.
-- no `_meta.*` tags are introduced for project information.
-- tutorials and accepted ADRs use `Workspace`.
diff --git a/docs/dev/benchmarking/20260519-103251_darwin-arm64_py314_tutorial-benchmarks.csv b/docs/dev/benchmarking/20260519-103251_darwin-arm64_py314_tutorial-benchmarks.csv
new file mode 100644
index 000000000..edb8e3f63
--- /dev/null
+++ b/docs/dev/benchmarking/20260519-103251_darwin-arm64_py314_tutorial-benchmarks.csv
@@ -0,0 +1,24 @@
+tutorial_name,elapsed_seconds,status,return_code
+ed-1.py,13.789,ok,0
+ed-10.py,39.098,ok,0
+ed-11.py,10.343,ok,0
+ed-12.py,8.331,ok,0
+ed-13.py,21.745,ok,0
+ed-14.py,6.158,ok,0
+ed-15.py,240.151,ok,0
+ed-16.py,58.481,ok,0
+ed-17.py,152.214,ok,0
+ed-18.py,6.114,ok,0
+ed-2.py,18.322,ok,0
+ed-20.py,36.422,ok,0
+ed-21.py,197.610,ok,0
+ed-22.py,194.351,ok,0
+ed-23.py,6.060,ok,0
+ed-24.py,4.749,ok,0
+ed-3.py,19.050,ok,0
+ed-4.py,4.480,ok,0
+ed-5.py,36.605,ok,0
+ed-6.py,61.846,ok,0
+ed-7.py,115.147,ok,0
+ed-8.py,101.442,ok,0
+ed-9.py,9.214,ok,0
diff --git a/docs/dev/benchmarking/20260519-103500_darwin-arm64_py314_tutorial-benchmarks.csv b/docs/dev/benchmarking/20260519-103500_darwin-arm64_py314_tutorial-benchmarks.csv
new file mode 100644
index 000000000..608944cb5
--- /dev/null
+++ b/docs/dev/benchmarking/20260519-103500_darwin-arm64_py314_tutorial-benchmarks.csv
@@ -0,0 +1,24 @@
+tutorial_name,elapsed_seconds,status,return_code
+ed-1.py,13.979,ok,0
+ed-10.py,38.764,ok,0
+ed-11.py,10.606,ok,0
+ed-12.py,9.044,ok,0
+ed-13.py,23.157,ok,0
+ed-14.py,6.585,ok,0
+ed-15.py,258.188,ok,0
+ed-16.py,60.097,ok,0
+ed-17.py,69.418,ok,0
+ed-18.py,6.181,ok,0
+ed-2.py,18.844,ok,0
+ed-20.py,39.157,ok,0
+ed-21.py,96.730,ok,0
+ed-22.py,73.480,ok,0
+ed-23.py,5.984,ok,0
+ed-24.py,4.942,ok,0
+ed-3.py,20.782,ok,0
+ed-4.py,5.780,ok,0
+ed-5.py,37.716,ok,0
+ed-6.py,66.911,ok,0
+ed-7.py,119.645,ok,0
+ed-8.py,103.887,ok,0
+ed-9.py,8.891,ok,0
diff --git a/docs/dev/benchmarking/20260519-121524_darwin-arm64_py314_tutorial-benchmarks.csv b/docs/dev/benchmarking/20260519-121524_darwin-arm64_py314_tutorial-benchmarks.csv
new file mode 100644
index 000000000..ff8eda19a
--- /dev/null
+++ b/docs/dev/benchmarking/20260519-121524_darwin-arm64_py314_tutorial-benchmarks.csv
@@ -0,0 +1,24 @@
+tutorial_name,elapsed_seconds,status,return_code
+ed-1.py,15.557,ok,0
+ed-10.py,40.860,ok,0
+ed-11.py,10.823,ok,0
+ed-12.py,8.861,ok,0
+ed-13.py,24.128,ok,0
+ed-14.py,6.722,ok,0
+ed-15.py,28.243,ok,0
+ed-16.py,59.218,ok,0
+ed-17.py,70.816,ok,0
+ed-18.py,6.944,ok,0
+ed-2.py,20.385,ok,0
+ed-20.py,39.513,ok,0
+ed-21.py,96.953,ok,0
+ed-22.py,75.390,ok,0
+ed-23.py,6.115,ok,0
+ed-24.py,5.159,ok,0
+ed-3.py,34.082,ok,0
+ed-4.py,8.215,ok,0
+ed-5.py,61.949,ok,0
+ed-6.py,83.857,ok,0
+ed-7.py,120.332,ok,0
+ed-8.py,103.831,ok,0
+ed-9.py,9.270,ok,0
diff --git a/docs/dev/issues/open.md b/docs/dev/issues/open.md
index 6c96f87ea..04d97ab85 100644
--- a/docs/dev/issues/open.md
+++ b/docs/dev/issues/open.md
@@ -1557,26 +1557,35 @@ threaded because each step's output is the next step's input.
 
 ---
 
-## 95. 🟡 Re-Enable DREAM Multiprocessing in CLI Workflows
+## 95. 🟡 Re-Enable DREAM Multiprocessing in Direct Python Scripts
 
-**Type:** Performance / CLI robustness
+**Type:** Performance / Script runtime
 
-On macOS and other spawn-based platforms, running Bayesian scripts via
-direct CLI entry points such as `python script.py` can fail during BUMPS
+On macOS and other spawn-based platforms, direct Bayesian tutorial
+execution via `python script.py` or wrappers such as
+`pixi run tutorial docs/docs/tutorials/ed-21.py` can fail during BUMPS
 `MPMapper` startup because worker processes re-import `__main__` and
 re-execute top-level tutorial code. The current defensive workaround is
 to fall back to serial execution for these direct-script entry points,
-which avoids the crash but disables DREAM multiprocessing in terminal
-workflows.
+which avoids the crash but disables DREAM multiprocessing and causes a
+large performance drop.
+
+Observed behavior for `ed-21` today:
+
+- Jupyter execution and `easydiffraction PROJECT_DIR fit` both appear to
+  use working parallel DREAM and complete `361/361` in about 40 seconds.
+- Direct Python-script execution of the same tutorial runs `361/361` in
+  about 220 seconds, consistent with the serial fallback path.
 
 **Possible solution:** keep the existing tracker-state cleanup before
 pickling and mapper startup, but replace the blanket serial fallback
 with an EasyDiffraction-controlled multiprocessing context policy. For
-direct CLI script entry points, prefer a `fork` context when available
-so workers do not re-import the tutorial top level. Keep the existing
-behavior for import-safe module entry points and for platforms where
-`fork` is unavailable. Document the tradeoff clearly because `fork` on
-macOS is less conservative than `spawn`.
+direct Python script entry points, prefer a `fork` context when
+available so workers do not re-import the tutorial top level. Keep the
+existing behavior for import-safe module entry points such as
+`easydiffraction PROJECT_DIR fit` and for platforms where `fork` is
+unavailable. Document the tradeoff clearly because `fork` on macOS is
+less conservative than `spawn`.
 
 **Depends on:** related to issue 89, but independent.
 
@@ -1688,90 +1697,90 @@ sampler progress displays — any fix should keep their visuals consistent
 
 ## Summary
 
-| #   | Issue                                            | Severity | Type                         |
-| --- | ------------------------------------------------ | -------- | ---------------------------- |
-| 3   | Rebuild joint-fit weights                        | 🟡 Med   | Fragility                    |
-| 5   | `Analysis` as `DatablockItem`                    | 🟡 Med   | Consistency                  |
-| 8   | Explicit `create()` signatures                   | 🟡 Med   | API safety                   |
-| 9   | Future enum extensions                           | 🟢 Low   | Design                       |
-| 10  | Unify update orchestration                       | 🟢 Low   | Maintainability              |
-| 11  | Document `_update` contract                      | 🟢 Low   | Maintainability              |
-| 13  | Suppress redundant dirty-flag sets               | 🟢 Low   | Performance                  |
-| 14  | Finer-grained change tracking                    | 🟢 Low   | Performance                  |
-| 15  | Validate joint-fit weights                       | 🟡 Med   | Correctness                  |
-| 17  | Use PDF-specific CIF names                       | 🟢 Low   | Naming                       |
-| 18  | Move CIF v2→v1 conversion out of calculator      | 🟢 Low   | Maintainability              |
-| 19  | Debug-mode logging for calculator imports        | 🟢 Low   | Diagnostics                  |
-| 20  | Redirect/suppress CrysPy stderr                  | 🟢 Low   | UX                           |
-| 21  | Clarify CrysPy TOF background CIF tags           | 🟡 Med   | Correctness                  |
-| 22  | Check SC instrument mapping in CrysPy            | 🟢 Low   | Correctness                  |
-| 23  | Investigate PyCrysFML pattern length discrepancy | 🟢 Low   | Correctness                  |
-| 24  | Process defaults on experiment creation          | 🟢 Low   | Design                       |
-| 25  | Refactor data `_update` methods                  | 🟡 Med   | Maintainability              |
-| 26  | Clarify `dtype` usage in data arrays             | 🟢 Low   | Cleanup                      |
-| 27  | Handle zero uncertainty in Bragg PD              | 🟢 Low   | Correctness                  |
-| 28  | Clarify Bragg PD data collection description     | 🟢 Low   | Cleanup                      |
-| 29  | Standardise CIF ID validator pattern             | 🟡 Med   | Consistency                  |
-| 30  | Make `refinement_status` default an Enum         | 🟢 Low   | Design                       |
-| 31  | Rename PD data point mixins                      | 🟢 Low   | Naming                       |
-| 32  | Move common methods to `DatablockCollection`     | 🟡 Med   | Maintainability              |
-| 33  | Make `_update_categories` abstract               | 🟡 Med   | Design                       |
-| 34  | Auto-extract `PeakProfileTypeEnum`               | 🟢 Low   | Design                       |
-| 35  | Rename `BeamModeEnum` members to CWL/TOF         | 🟢 Low   | Naming                       |
-| 36  | Common `EnumBase` class                          | 🟢 Low   | Design                       |
-| 37  | Rename experiment `.type` property               | 🟢 Low   | Naming                       |
-| 38  | Fix `@typechecked`/gemmi in factories            | 🟡 Med   | Bug                          |
-| 39  | Improve `_update_priority` handling              | 🟢 Low   | Design                       |
-| 40  | Reset `.user_constrained` to `False`             | 🟢 Low   | Feature                      |
-| 41  | Check `_mark_dirty` in `_set_value`              | 🟢 Low   | Cleanup                      |
-| 42  | MkDocs type unpacking in validation              | 🟢 Low   | Docs                         |
-| 43  | Fix summary display inconsistencies              | 🟢 Low   | UX                           |
-| 44  | Merge parameter record construction              | 🟢 Low   | Cleanup                      |
-| 45  | Decide alias/constraint descriptor default       | 🟢 Low   | Design                       |
-| 46  | Improve `JointFitItem` descriptions              | 🟢 Low   | Naming                       |
-| 47  | Improve error handling in crystallography        | 🟢 Low   | Diagnostics                  |
-| 48  | Fix CrysPy TOF instrument default                | 🟢 Low   | Bug workaround               |
-| 49  | Automate space group CIF name variants           | 🟢 Low   | Maintainability              |
-| 50  | Clarify `Cell._update` minimizer param           | 🟢 Low   | Cleanup                      |
-| 51  | Access space group for Wyckoff letters           | 🟢 Low   | Design                       |
-| 52  | Rename line-segment `y` to `intensity`           | 🟢 Low   | Naming                       |
-| 53  | Move `show()` to `CategoryCollection`            | 🟢 Low   | Maintainability              |
-| 54  | Add `point_id` to excluded regions               | 🟢 Low   | Completeness                 |
-| 55  | Fix Jupyter scroll disabling for MkDocs          | 🟢 Low   | Docs / UX                    |
-| 56  | Make ASCII plot width configurable               | 🟢 Low   | UX                           |
-| 57  | Clean up CIF deserialisation helpers             | 🟢 Low   | Maintainability              |
-| 58  | Move `ProjectInfo` CIF methods to `serialize`    | 🟢 Low   | Maintainability              |
-| 59  | Add CIF name validation in parse                 | 🟢 Low   | Robustness                   |
-| 60  | Unify `mkdir` usage                              | 🟢 Low   | Cleanup                      |
-| 61  | Clarify logger default reaction mode             | 🟢 Low   | Design                       |
-| 62  | Complete `render_table` → `TableRenderer`        | 🟢 Low   | Cleanup                      |
-| 63  | Fix calculator `calculate_pattern` signature     | 🟢 Low   | Design                       |
-| 64  | Check unused-if-loading-from-CIF code            | 🟢 Low   | Cleanup                      |
-| 65  | Replace all bare `print()` with logging          | 🟡 Med   | Code quality                 |
-| 66  | Error-handling strategy: `log.error` vs `raise`  | 🟡 Med   | Design                       |
-| 67  | Custom validation for params and category types  | 🟡 Med   | Design                       |
-| 68  | `@typechecked` on all public methods?            | 🟢 Low   | Design                       |
-| 69  | Shorter public API names via `__init__`          | 🟢 Low   | API ergonomics               |
-| 70  | Standardise class member ordering + headers      | 🟡 Med   | Code style                   |
-| 71  | `_update_priority` reference table               | 🟢 Low   | Documentation                |
-| 72  | Warn on all switchable-category type changes     | 🟡 Med   | UX                           |
-| 73  | Unify setter parameter naming                    | 🟢 Low   | Code style                   |
-| 74  | Sync property type hints + custom lint rules     | 🟡 Med   | Tooling                      |
-| 75  | `show_supported_calculators()` on Analysis       | 🟢 Low   | API completeness             |
-| 76  | Consistent `_type` suffix in switchable APIs     | 🟡 Med   | Naming                       |
-| 79  | Verify analysis CIF serialisation completeness   | 🟢 Low   | Correctness                  |
-| 80  | Resolve `Any` vs `object` annotation policy      | 🟢 Low   | Code style                   |
-| 81  | Enforce docstrings on all public methods         | 🟡 Med   | Code quality                 |
-| 82  | Document `param-docstring-fix` workflow          | 🟢 Low   | Documentation                |
-| 83  | Remove redundant parameter listing               | 🟢 Low   | Cleanup                      |
-| 84  | Serialise `None` as `.` in CIF output            | 🟡 Med   | Correctness                  |
-| 85  | Retain per-experiment fitted params for plotting | 🟡 Med   | Correctness                  |
-| 86  | Auto-resolve `plot_param` x-axis + add units     | 🟢 Low   | UX                           |
-| 87  | Redesign tutorial grouping/categorisation        | 🟢 Low   | Documentation                |
-| 88  | Fix Dataset 26 description (47 not 57)           | 🟢 Low   | Data                         |
-| 89  | Parallel independent fits for single mode        | 🟡 Med   | Performance                  |
-| 95  | Re-enable DREAM multiprocessing in CLI workflows | 🟡 Med   | Performance / CLI robustness |
-| 90  | Show experiment number during sequential fitting | 🟢 Low   | UX                           |
-| 91  | Disable TODO checks in CodeFactor PRs            | 🟢 Low   | CI / Tooling                 |
-| 92  | Make `save()` respect verbosity                  | 🟢 Low   | UX                           |
-| 93  | Eliminate flicker in live progress tables        | 🟡 Med   | UX                           |
+| #   | Issue                                             | Severity | Type                         |
+| --- | ------------------------------------------------- | -------- | ---------------------------- |
+| 3   | Rebuild joint-fit weights                         | 🟡 Med   | Fragility                    |
+| 5   | `Analysis` as `DatablockItem`                     | 🟡 Med   | Consistency                  |
+| 8   | Explicit `create()` signatures                    | 🟡 Med   | API safety                   |
+| 9   | Future enum extensions                            | 🟢 Low   | Design                       |
+| 10  | Unify update orchestration                        | 🟢 Low   | Maintainability              |
+| 11  | Document `_update` contract                       | 🟢 Low   | Maintainability              |
+| 13  | Suppress redundant dirty-flag sets                | 🟢 Low   | Performance                  |
+| 14  | Finer-grained change tracking                     | 🟢 Low   | Performance                  |
+| 15  | Validate joint-fit weights                        | 🟡 Med   | Correctness                  |
+| 17  | Use PDF-specific CIF names                        | 🟢 Low   | Naming                       |
+| 18  | Move CIF v2→v1 conversion out of calculator       | 🟢 Low   | Maintainability              |
+| 19  | Debug-mode logging for calculator imports         | 🟢 Low   | Diagnostics                  |
+| 20  | Redirect/suppress CrysPy stderr                   | 🟢 Low   | UX                           |
+| 21  | Clarify CrysPy TOF background CIF tags            | 🟡 Med   | Correctness                  |
+| 22  | Check SC instrument mapping in CrysPy             | 🟢 Low   | Correctness                  |
+| 23  | Investigate PyCrysFML pattern length discrepancy  | 🟢 Low   | Correctness                  |
+| 24  | Process defaults on experiment creation           | 🟢 Low   | Design                       |
+| 25  | Refactor data `_update` methods                   | 🟡 Med   | Maintainability              |
+| 26  | Clarify `dtype` usage in data arrays              | 🟢 Low   | Cleanup                      |
+| 27  | Handle zero uncertainty in Bragg PD               | 🟢 Low   | Correctness                  |
+| 28  | Clarify Bragg PD data collection description      | 🟢 Low   | Cleanup                      |
+| 29  | Standardise CIF ID validator pattern              | 🟡 Med   | Consistency                  |
+| 30  | Make `refinement_status` default an Enum          | 🟢 Low   | Design                       |
+| 31  | Rename PD data point mixins                       | 🟢 Low   | Naming                       |
+| 32  | Move common methods to `DatablockCollection`      | 🟡 Med   | Maintainability              |
+| 33  | Make `_update_categories` abstract                | 🟡 Med   | Design                       |
+| 34  | Auto-extract `PeakProfileTypeEnum`                | 🟢 Low   | Design                       |
+| 35  | Rename `BeamModeEnum` members to CWL/TOF          | 🟢 Low   | Naming                       |
+| 36  | Common `EnumBase` class                           | 🟢 Low   | Design                       |
+| 37  | Rename experiment `.type` property                | 🟢 Low   | Naming                       |
+| 38  | Fix `@typechecked`/gemmi in factories             | 🟡 Med   | Bug                          |
+| 39  | Improve `_update_priority` handling               | 🟢 Low   | Design                       |
+| 40  | Reset `.user_constrained` to `False`              | 🟢 Low   | Feature                      |
+| 41  | Check `_mark_dirty` in `_set_value`               | 🟢 Low   | Cleanup                      |
+| 42  | MkDocs type unpacking in validation               | 🟢 Low   | Docs                         |
+| 43  | Fix summary display inconsistencies               | 🟢 Low   | UX                           |
+| 44  | Merge parameter record construction               | 🟢 Low   | Cleanup                      |
+| 45  | Decide alias/constraint descriptor default        | 🟢 Low   | Design                       |
+| 46  | Improve `JointFitItem` descriptions               | 🟢 Low   | Naming                       |
+| 47  | Improve error handling in crystallography         | 🟢 Low   | Diagnostics                  |
+| 48  | Fix CrysPy TOF instrument default                 | 🟢 Low   | Bug workaround               |
+| 49  | Automate space group CIF name variants            | 🟢 Low   | Maintainability              |
+| 50  | Clarify `Cell._update` minimizer param            | 🟢 Low   | Cleanup                      |
+| 51  | Access space group for Wyckoff letters            | 🟢 Low   | Design                       |
+| 52  | Rename line-segment `y` to `intensity`            | 🟢 Low   | Naming                       |
+| 53  | Move `show()` to `CategoryCollection`             | 🟢 Low   | Maintainability              |
+| 54  | Add `point_id` to excluded regions                | 🟢 Low   | Completeness                 |
+| 55  | Fix Jupyter scroll disabling for MkDocs           | 🟢 Low   | Docs / UX                    |
+| 56  | Make ASCII plot width configurable                | 🟢 Low   | UX                           |
+| 57  | Clean up CIF deserialisation helpers              | 🟢 Low   | Maintainability              |
+| 58  | Move `ProjectInfo` CIF methods to `serialize`     | 🟢 Low   | Maintainability              |
+| 59  | Add CIF name validation in parse                  | 🟢 Low   | Robustness                   |
+| 60  | Unify `mkdir` usage                               | 🟢 Low   | Cleanup                      |
+| 61  | Clarify logger default reaction mode              | 🟢 Low   | Design                       |
+| 62  | Complete `render_table` → `TableRenderer`         | 🟢 Low   | Cleanup                      |
+| 63  | Fix calculator `calculate_pattern` signature      | 🟢 Low   | Design                       |
+| 64  | Check unused-if-loading-from-CIF code             | 🟢 Low   | Cleanup                      |
+| 65  | Replace all bare `print()` with logging           | 🟡 Med   | Code quality                 |
+| 66  | Error-handling strategy: `log.error` vs `raise`   | 🟡 Med   | Design                       |
+| 67  | Custom validation for params and category types   | 🟡 Med   | Design                       |
+| 68  | `@typechecked` on all public methods?             | 🟢 Low   | Design                       |
+| 69  | Shorter public API names via `__init__`           | 🟢 Low   | API ergonomics               |
+| 70  | Standardise class member ordering + headers       | 🟡 Med   | Code style                   |
+| 71  | `_update_priority` reference table                | 🟢 Low   | Documentation                |
+| 72  | Warn on all switchable-category type changes      | 🟡 Med   | UX                           |
+| 73  | Unify setter parameter naming                     | 🟢 Low   | Code style                   |
+| 74  | Sync property type hints + custom lint rules      | 🟡 Med   | Tooling                      |
+| 75  | `show_supported_calculators()` on Analysis        | 🟢 Low   | API completeness             |
+| 76  | Consistent `_type` suffix in switchable APIs      | 🟡 Med   | Naming                       |
+| 79  | Verify analysis CIF serialisation completeness    | 🟢 Low   | Correctness                  |
+| 80  | Resolve `Any` vs `object` annotation policy       | 🟢 Low   | Code style                   |
+| 81  | Enforce docstrings on all public methods          | 🟡 Med   | Code quality                 |
+| 82  | Document `param-docstring-fix` workflow           | 🟢 Low   | Documentation                |
+| 83  | Remove redundant parameter listing                | 🟢 Low   | Cleanup                      |
+| 84  | Serialise `None` as `.` in CIF output             | 🟡 Med   | Correctness                  |
+| 85  | Retain per-experiment fitted params for plotting  | 🟡 Med   | Correctness                  |
+| 86  | Auto-resolve `plot_param` x-axis + add units      | 🟢 Low   | UX                           |
+| 87  | Redesign tutorial grouping/categorisation         | 🟢 Low   | Documentation                |
+| 88  | Fix Dataset 26 description (47 not 57)            | 🟢 Low   | Data                         |
+| 89  | Parallel independent fits for single mode         | 🟡 Med   | Performance                  |
+| 95  | Re-enable DREAM multiprocessing in direct scripts | 🟡 Med   | Performance / Script runtime |
+| 90  | Show experiment number during sequential fitting  | 🟢 Low   | UX                           |
+| 91  | Disable TODO checks in CodeFactor PRs             | 🟢 Low   | CI / Tooling                 |
+| 92  | Make `save()` respect verbosity                   | 🟢 Low   | UX                           |
+| 93  | Eliminate flicker in live progress tables         | 🟡 Med   | UX                           |
diff --git a/docs/dev/package-structure/full.md b/docs/dev/package-structure/full.md
index 720505baa..677f5851d 100644
--- a/docs/dev/package-structure/full.md
+++ b/docs/dev/package-structure/full.md
@@ -24,6 +24,54 @@
 │   │   │   │   └── 🏷️ class Aliases
 │   │   │   └── 📄 factory.py
 │   │   │       └── 🏷️ class AliasesFactory
+│   │   ├── 📁 bayesian_convergence
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   └── 🏷️ class BayesianConvergence
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class BayesianConvergenceFactory
+│   │   ├── 📁 bayesian_distribution_caches
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   ├── 🏷️ class BayesianDistributionCacheItem
+│   │   │   │   └── 🏷️ class BayesianDistributionCaches
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class BayesianDistributionCachesFactory
+│   │   ├── 📁 bayesian_pair_caches
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   ├── 🏷️ class BayesianPairCachePaths
+│   │   │   │   ├── 🏷️ class BayesianPairCacheItem
+│   │   │   │   └── 🏷️ class BayesianPairCaches
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class BayesianPairCachesFactory
+│   │   ├── 📁 bayesian_parameter_posteriors
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   ├── 🏷️ class BayesianParameterPosteriorItem
+│   │   │   │   └── 🏷️ class BayesianParameterPosteriors
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class BayesianParameterPosteriorsFactory
+│   │   ├── 📁 bayesian_predictive_datasets
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   ├── 🏷️ class BayesianPredictiveDatasetPaths
+│   │   │   │   ├── 🏷️ class BayesianPredictiveDatasetItem
+│   │   │   │   └── 🏷️ class BayesianPredictiveDatasets
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class BayesianPredictiveDatasetsFactory
+│   │   ├── 📁 bayesian_result
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   └── 🏷️ class BayesianResult
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class BayesianResultFactory
+│   │   ├── 📁 bayesian_sampler
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   └── 🏷️ class BayesianSampler
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class BayesianSamplerFactory
 │   │   ├── 📁 constraints
 │   │   │   ├── 📄 __init__.py
 │   │   │   ├── 📄 default.py
@@ -31,6 +79,33 @@
 │   │   │   │   └── 🏷️ class Constraints
 │   │   │   └── 📄 factory.py
 │   │   │       └── 🏷️ class ConstraintsFactory
+│   │   ├── 📁 deterministic_result
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   └── 🏷️ class DeterministicResult
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class DeterministicResultFactory
+│   │   ├── 📁 fit_parameter_correlations
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   ├── 🏷️ class FitParameterCorrelationItem
+│   │   │   │   └── 🏷️ class FitParameterCorrelations
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class FitParameterCorrelationsFactory
+│   │   ├── 📁 fit_parameters
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   ├── 🏷️ class FitParameterItem
+│   │   │   │   └── 🏷️ class FitParameters
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class FitParametersFactory
+│   │   ├── 📁 fit_result
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   └── 🏷️ class FitResult
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class FitResultFactory
+│   │   ├── 📁 fit_state
 │   │   ├── 📁 fitting
 │   │   │   ├── 📄 __init__.py
 │   │   │   ├── 📄 default.py
@@ -76,7 +151,9 @@
 │   │   ├── 📄 base.py
 │   │   │   └── 🏷️ class MinimizerBase
 │   │   ├── 📄 bumps.py
+│   │   │   ├── 🏷️ class _BumpsEvaluationLimitError
 │   │   │   ├── 🏷️ class _EasyDiffractionFitness
+│   │   │   ├── 🏷️ class _BumpsProgressMonitor
 │   │   │   └── 🏷️ class BumpsMinimizer
 │   │   ├── 📄 bumps_amoeba.py
 │   │   │   └── 🏷️ class BumpsAmoebaMinimizer
@@ -105,9 +182,13 @@
 │   ├── 📄 __init__.py
 │   ├── 📄 analysis.py
 │   │   ├── 🏷️ class AnalysisDisplay
+│   │   ├── 🏷️ class _AnalysisOwnerAccessorsMixin
+│   │   ├── 🏷️ class _AnalysisPersistedCategoryAccessorsMixin
 │   │   └── 🏷️ class Analysis
 │   ├── 📄 enums.py
-│   │   └── 🏷️ class FitModeEnum
+│   │   ├── 🏷️ class FitModeEnum
+│   │   ├── 🏷️ class FitResultKindEnum
+│   │   └── 🏷️ class FitCorrelationSourceEnum
 │   ├── 📄 fitting.py
 │   │   └── 🏷️ class Fitter
 │   └── 📄 sequential.py
@@ -159,10 +240,12 @@
 │       ├── 🏷️ class GenericStringDescriptor
 │       ├── 🏷️ class GenericBoolDescriptor
 │       ├── 🏷️ class GenericNumericDescriptor
+│       ├── 🏷️ class GenericIntegerDescriptor
 │       ├── 🏷️ class GenericParameter
 │       ├── 🏷️ class StringDescriptor
 │       ├── 🏷️ class BoolDescriptor
 │       ├── 🏷️ class NumericDescriptor
+│       ├── 🏷️ class IntegerDescriptor
 │       └── 🏷️ class Parameter
 ├── 📁 crystallography
 │   ├── 📄 __init__.py
@@ -424,7 +507,8 @@
 │   │   ├── 📄 parse.py
 │   │   └── 📄 serialize.py
 │   ├── 📄 __init__.py
-│   └── 📄 ascii.py
+│   ├── 📄 ascii.py
+│   └── 📄 results_sidecar.py
 ├── 📁 project
 │   ├── 📁 categories
 │   │   ├── 📁 info
@@ -439,6 +523,12 @@
 │   │   │   │   └── 🏷️ class Rendering
 │   │   │   └── 📄 factory.py
 │   │   │       └── 🏷️ class RenderingFactory
+│   │   ├── 📁 verbosity
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   │   └── 🏷️ class Verbosity
+│   │   │   └── 📄 factory.py
+│   │   │       └── 🏷️ class VerbosityFactory
 │   │   └── 📄 __init__.py
 │   ├── 📄 __init__.py
 │   ├── 📄 display.py
diff --git a/docs/dev/package-structure/short.md b/docs/dev/package-structure/short.md
index da42f7a57..367a36a51 100644
--- a/docs/dev/package-structure/short.md
+++ b/docs/dev/package-structure/short.md
@@ -15,10 +15,55 @@
 │   │   │   ├── 📄 __init__.py
 │   │   │   ├── 📄 default.py
 │   │   │   └── 📄 factory.py
+│   │   ├── 📁 bayesian_convergence
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 bayesian_distribution_caches
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 bayesian_pair_caches
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 bayesian_parameter_posteriors
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 bayesian_predictive_datasets
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 bayesian_result
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 bayesian_sampler
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
 │   │   ├── 📁 constraints
 │   │   │   ├── 📄 __init__.py
 │   │   │   ├── 📄 default.py
 │   │   │   └── 📄 factory.py
+│   │   ├── 📁 deterministic_result
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 fit_parameter_correlations
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 fit_parameters
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 fit_result
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
+│   │   ├── 📁 fit_state
 │   │   ├── 📁 fitting
 │   │   │   ├── 📄 __init__.py
 │   │   │   ├── 📄 default.py
@@ -205,7 +250,8 @@
 │   │   ├── 📄 parse.py
 │   │   └── 📄 serialize.py
 │   ├── 📄 __init__.py
-│   └── 📄 ascii.py
+│   ├── 📄 ascii.py
+│   └── 📄 results_sidecar.py
 ├── 📁 project
 │   ├── 📁 categories
 │   │   ├── 📁 info
@@ -216,6 +262,10 @@
 │   │   │   ├── 📄 __init__.py
 │   │   │   ├── 📄 default.py
 │   │   │   └── 📄 factory.py
+│   │   ├── 📁 verbosity
+│   │   │   ├── 📄 __init__.py
+│   │   │   ├── 📄 default.py
+│   │   │   └── 📄 factory.py
 │   │   └── 📄 __init__.py
 │   ├── 📄 __init__.py
 │   ├── 📄 display.py
diff --git a/docs/dev/plans/loop-category-key-identity.md b/docs/dev/plans/loop-category-key-identity.md
deleted file mode 100644
index 9917d1bae..000000000
--- a/docs/dev/plans/loop-category-key-identity.md
+++ /dev/null
@@ -1,620 +0,0 @@
-# Loop Category Key Identity Implementation Plan
-
-## Status
-
-Workflow instructions:
-
-```text
-.github/copilot-instructions.md
-```
-
-Related ADR suggestion:
-
-```text
-docs/dev/adrs/suggestions/loop-category-key-identity.md
-```
-
-This plan implements two related changes:
-
-1. Move category identity declarations from per-instance assignments to
-   class-level declarations.
-2. Add an explicit persisted `id` field to the constraints loop.
-
-Status checklist. Mark `[x]` only while implementing:
-
-```text
-Phase 1 - Implementation
-[x] Add class-level identity declarations to CategoryItem.
-[x] Teach Identity to resolve declared category entry names.
-[x] Rebuild collection indexes after CIF loop loading.
-[x] Add _category_code to all current CategoryItem subclasses.
-[x] Add _category_entry_name to all current loop CategoryItem subclasses.
-[x] Remove direct self._identity.category_code assignments.
-[x] Remove direct self._identity.category_entry_name lambda assignments.
-[x] Add Constraint.id descriptor serialized as _constraint.id.
-[x] Change constraints collection keys from lhs_alias to id.
-[x] Preserve default constraints.create(expression=...) behavior by using lhs_alias as the default id.
-[x] Add backward-compatible loading for old CIF loops without _constraint.id.
-[x] Update constraints display.
-[x] Update loop-category-key-identity.md if implementation details differ from the ADR.
-[ ] Phase 1 review gate: present diff for approval.
-
-Phase 2 - Verification
-[x] Add tests for the base declarative identity behavior.
-[x] Add parametrized tests for current loop category identity declarations.
-[x] Update constraints tests.
-[x] Update existing round-trip tests that compare constraints CIF.
-[x] Run formatting.
-[x] Run targeted unit tests.
-[x] Run broader checks.
-```
-
-## Commit Discipline
-
-When an AI agent follows this plan, every completed Phase 1
-implementation step must be staged with explicit paths and committed
-locally before moving to the next implementation step or to the Phase 1
-review gate.
-
-Follow the **Commits** section of `.github/copilot-instructions.md`.
-
-Rules:
-
-- One commit per implementation step.
-- Keep each commit atomic and single-purpose.
-- Stage explicit paths only. Do not use `git add .`.
-- Do not stage unrelated user changes.
-- Do not stage generated artifacts unless the user explicitly asks.
-- If a serious uncovered design issue appears, stop and ask before
-  continuing.
-
-Suggested branch:
-
-```text
-feature/loop-category-key-identity
-```
-
-Suggested commit messages:
-
-```text
-Add declarative category identity resolution
-Declare category identities on current items
-Persist explicit constraint identifiers
-Update ADR for declarative category identity
-Add declarative category identity tests
-```
-
-## Goal
-
-Replace repeated constructor code like this:
-
-```python
-self._identity.category_code = 'atom_site'
-self._identity.category_entry_name = lambda: str(self.label.value)
-```
-
-with class-level declarations:
-
-```python
-class AtomSite(CategoryItem):
-    _category_code = 'atom_site'
-    _category_entry_name = 'label'
-```
-
-The name `_category_entry_name` is intentionally kept because this is
-the preferred project terminology. In this plan it means "the name of
-the item attribute used to resolve the entry name". The resolved entry
-value is still obtained through:
-
-```python
-item._identity.category_entry_name
-```
-
-For example, `AtomSite._category_entry_name == 'label'`, while
-`atom_site._identity.category_entry_name == 'Ba1'`.
-
-## Non-Goals
-
-Do not change these in this migration:
-
-- Do not rename `Identity.category_entry_name`.
-- Do not rename `CategoryCollection`.
-- Do not change public collection access syntax.
-- Do not change CIF tags except adding `_constraint.id`.
-- Do not rename `label` to `id` for atom sites or aliases.
-- Do not make `phase_id` the row key for powder reflection loops.
-
-## Design
-
-This plan intentionally uses a narrow metadata lookup based on
-`_category_entry_name`. That is an allowed exception to the general "no
-string-based dispatch" rule because the attribute name is a class-level
-declaration, not user input, and resolution is centralized in
-`CategoryItem`.
-
-### CategoryItem Declarations
-
-Add these class attributes to `CategoryItem` in
-`src/easydiffraction/core/category.py`:
-
-```python
-class CategoryItem(GuardedBase):
-    _category_code: str | None = None
-    _category_entry_name: str | None = None
-```
-
-Update `CategoryItem.__init__()` so it assigns `_category_code` once:
-
-```python
-def __init__(self) -> None:
-    super().__init__()
-    if self._category_code is not None:
-        self._identity.category_code = self._category_code
-```
-
-Do not try to resolve `_category_entry_name` in `CategoryItem.__init__`.
-Many item classes create descriptors after `super().__init__()` returns,
-and some mixin-based classes run `CategoryItem.__init__()` before their
-descriptors are created.
-
-Add a resolver method to `CategoryItem`:
-
-```python
-def _resolve_category_entry_name(self) -> str | None:
-    attr_name = self._category_entry_name
-    if attr_name is None:
-        return None
-
-    value = getattr(self, attr_name)
-    if isinstance(value, GenericDescriptorBase):
-        value = value.value
-    return str(value)
-```
-
-Import `GenericDescriptorBase` from `easydiffraction.core.variable` in
-`category.py` if it is not already in scope.
-
-### Identity Resolution
-
-Update `Identity._resolve_up()` in
-`src/easydiffraction/core/identity.py` so that category entries can be
-resolved from the owning object before walking to the parent.
-
-Add this logic after checking direct callable/string values and before
-climbing to the parent:
-
-```python
-if attr == 'category_entry':
-    resolver = getattr(self._owner, '_resolve_category_entry_name', None)
-    if callable(resolver):
-        resolved = resolver()
-        if resolved is not None:
-            return resolved
-```
-
-Keep the existing `category_entry_name` setter. It remains useful as an
-escape hatch and keeps old code compatible during migration.
-
-### Collection Index Rebuild After CIF Loading
-
-Update `category_collection_from_cif()` in
-`src/easydiffraction/io/cif/serialize.py`.
-
-Currently `_adopt_items()` rebuilds the index before loop values are
-loaded into each item. After declarative keys are resolved from
-descriptor values, the index must be rebuilt after parameters are
-loaded.
-
-After the row population loop, run any collection hook and rebuild:
-
-```python
-after_from_cif = getattr(self, '_after_from_cif', None)
-if callable(after_from_cif):
-    after_from_cif()
-
-self._rebuild_index()
-```
-
-The hook is needed for constraints to backfill missing ids from old CIF
-files.
-
-## Category Migration Table
-
-Add `_category_code` to each listed class. Add `_category_entry_name`
-only when the class is a loop item and currently has a collection key.
-Then remove matching constructor assignments.
-
-| File                                                                                | Class                      | `_category_code`         | `_category_entry_name` | Notes                                        |
-| ----------------------------------------------------------------------------------- | -------------------------- | ------------------------ | ---------------------- | -------------------------------------------- |
-| `src/easydiffraction/project/categories/info/default.py`                            | `ProjectInfo`              | `project`                | none                   | Singleton category.                          |
-| `src/easydiffraction/project/categories/rendering/default.py`                       | `Rendering`                | `rendering`              | none                   | Singleton category.                          |
-| `src/easydiffraction/analysis/categories/fitting/default.py`                        | `Fitting`                  | `fitting`                | none                   | Singleton category.                          |
-| `src/easydiffraction/analysis/categories/sequential_fit/default.py`                 | `SequentialFit`            | `sequential_fit`         | none                   | Singleton category.                          |
-| `src/easydiffraction/analysis/categories/aliases/default.py`                        | `Alias`                    | `alias`                  | `label`                | Loop key stays `_alias.label`.               |
-| `src/easydiffraction/analysis/categories/constraints/default.py`                    | `Constraint`               | `constraint`             | `id`                   | Add the id descriptor first.                 |
-| `src/easydiffraction/analysis/categories/joint_fit/default.py`                      | `JointFitItem`             | `joint_fit`              | `experiment_id`        | Loop key stays `_joint_fit.experiment_id`.   |
-| `src/easydiffraction/analysis/categories/sequential_fit_extract/default.py`         | `SequentialFitExtractItem` | `sequential_fit_extract` | `id`                   | Loop key stays `_sequential_fit_extract.id`. |
-| `src/easydiffraction/datablocks/structure/categories/cell/default.py`               | `Cell`                     | `cell`                   | none                   | Singleton category.                          |
-| `src/easydiffraction/datablocks/structure/categories/space_group/default.py`        | `SpaceGroup`               | `space_group`            | none                   | Singleton category.                          |
-| `src/easydiffraction/datablocks/structure/categories/atom_sites/default.py`         | `AtomSite`                 | `atom_site`              | `label`                | Loop key stays `_atom_site.label`.           |
-| `src/easydiffraction/datablocks/structure/categories/atom_site_aniso/default.py`    | `AtomSiteAniso`            | `atom_site_aniso`        | `label`                | Loop key stays `_atom_site_aniso.label`.     |
-| `src/easydiffraction/datablocks/experiment/categories/experiment_type/default.py`   | `ExperimentType`           | `expt_type`              | none                   | Singleton category.                          |
-| `src/easydiffraction/datablocks/experiment/categories/calculation/default.py`       | `Calculation`              | `calculation`            | none                   | Singleton category.                          |
-| `src/easydiffraction/datablocks/experiment/categories/diffrn/default.py`            | `DefaultDiffrn`            | `diffrn`                 | none                   | Singleton category.                          |
-| `src/easydiffraction/datablocks/experiment/categories/instrument/base.py`           | `InstrumentBase`           | `instrument`             | none                   | Subclasses inherit this code.                |
-| `src/easydiffraction/datablocks/experiment/categories/peak/base.py`                 | `PeakBase`                 | `peak`                   | none                   | Subclasses inherit this code.                |
-| `src/easydiffraction/datablocks/experiment/categories/extinction/becker_coppens.py` | `BeckerCoppensExtinction`  | `extinction`             | none                   | Singleton category.                          |
-| `src/easydiffraction/datablocks/experiment/categories/linked_crystal/default.py`    | `LinkedCrystal`            | `linked_crystal`         | none                   | Singleton category.                          |
-| `src/easydiffraction/datablocks/experiment/categories/linked_phases/default.py`     | `LinkedPhase`              | `linked_phases`          | `id`                   | Loop key stays `_pd_phase_block.id`.         |
-| `src/easydiffraction/datablocks/experiment/categories/background/line_segment.py`   | `LineSegment`              | `background`             | `id`                   | Loop key stays `_pd_background.id`.          |
-| `src/easydiffraction/datablocks/experiment/categories/background/chebyshev.py`      | `PolynomialTerm`           | `background`             | `id`                   | Loop key stays `_pd_background.id`.          |
-| `src/easydiffraction/datablocks/experiment/categories/excluded_regions/default.py`  | `ExcludedRegion`           | `excluded_regions`       | `id`                   | Loop key stays `_excluded_region.id`.        |
-| `src/easydiffraction/datablocks/experiment/categories/refln/bragg_sc.py`            | `Refln`                    | `refln`                  | `id`                   | Powder reflection rows inherit this key.     |
-| `src/easydiffraction/datablocks/experiment/categories/data/bragg_pd.py`             | `PdCwlDataPoint`           | `pd_data`                | `point_id`             | CWL powder data row.                         |
-| `src/easydiffraction/datablocks/experiment/categories/data/bragg_pd.py`             | `PdTofDataPoint`           | `pd_data`                | `point_id`             | TOF powder data row.                         |
-| `src/easydiffraction/datablocks/experiment/categories/data/total_pd.py`             | `TotalDataPoint`           | `total_data`             | `point_id`             | Total-scattering data row.                   |
-
-Do not add `_category_code` or `_category_entry_name` to
-`PowderReflnBase`, `PowderCwlRefln`, or `PowderTofRefln`; they inherit
-the `refln` identity declarations from `Refln`.
-
-## Constraints Migration
-
-### Target Shape
-
-`Constraint` should become:
-
-```python
-class Constraint(CategoryItem):
-    _category_code = 'constraint'
-    _category_entry_name = 'id'
-
-    def __init__(self) -> None:
-        super().__init__()
-
-        self._id = StringDescriptor(
-            name='id',
-            description='Identifier for this constraint.',
-            value_spec=AttributeSpec(
-                default='_',
-                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_]*$'),
-            ),
-            cif_handler=CifHandler(names=['_constraint.id']),
-        )
-        self._expression = StringDescriptor(...)
-```
-
-Define the public property:
-
-```python
-@property
-def id(self) -> StringDescriptor:
-    """Identifier for this constraint."""
-    return self._id
-
-@id.setter
-def id(self, value: str) -> None:
-    self._id.value = value
-```
-
-Keep `lhs_alias` and `rhs_expr` as derived read-only properties.
-
-### Create API
-
-Change `Constraints.create()` from:
-
-```python
-def create(self, *, expression: str) -> None:
-```
-
-to:
-
-```python
-def create(self, *, expression: str, id: str | None = None) -> None:
-```
-
-Implementation order:
-
-```python
-item = Constraint()
-item.expression = expression
-item.id = id if id is not None else item.lhs_alias
-self.add(item)
-self._enabled = True
-```
-
-This preserves the current default user experience:
-
-```python
-analysis.constraints.create(expression='biso_Ba = biso_La')
-analysis.constraints['biso_Ba']
-```
-
-It also allows explicit ids:
-
-```python
-analysis.constraints.create(
-    id='constraint_1',
-    expression='biso_Ba = biso_La',
-)
-analysis.constraints['constraint_1']
-```
-
-### Backward-Compatible CIF Loading
-
-Old CIF files only contain:
-
-```cif
-loop_
-_constraint.expression
-biso_Ba = biso_La
-```
-
-New CIF files should contain:
-
-```cif
-loop_
-_constraint.id
-_constraint.expression
-biso_Ba "biso_Ba = biso_La"
-```
-
-Add a hook on `Constraints`:
-
-```python
-def _after_from_cif(self) -> None:
-    for index, item in enumerate(self._items, start=1):
-        if item.id.value in {'', '_'}:
-            fallback = item.lhs_alias or f'constraint_{index}'
-            item.id = fallback
-```
-
-The generic `category_collection_from_cif()` hook described above must
-call this before rebuilding the index.
-
-### Constraints Display
-
-Update `Constraints.show()` to include the id:
-
-```text
-id | expression
-```
-
-Keep the existing empty warning behavior.
-
-## Required Code Searches
-
-After migration, these searches should return no category-item
-constructor assignments:
-
-```shell
-git grep -n -E "self\\._identity\\.category_code =" -- src/easydiffraction
-git grep -n -E "self\\._identity\\.category_entry_name =" -- src/easydiffraction
-git grep -n -E "category_entry_name = lambda" -- src/easydiffraction
-```
-
-The following reads are expected to remain:
-
-```shell
-git grep -n "_identity\\.category_entry_name" -- src/easydiffraction
-```
-
-Those reads are used by collections, display, reporting, and parameter
-unique names.
-
-Do not use `rg` in this plan; it is not available in every contributor
-environment. Use `git grep` for repository searches.
-
-## Tests To Add Or Update
-
-### Base Identity Tests
-
-Add or update tests under `tests/unit/easydiffraction/core/`.
-
-Test a fake category item:
-
-```python
-class FakeItem(CategoryItem):
-    _category_code = 'fake'
-    _category_entry_name = 'id'
-
-    def __init__(self):
-        super().__init__()
-        self._id = StringDescriptor(...)
-
-    @property
-    def id(self):
-        return self._id
-```
-
-Assert:
-
-```python
-item._identity.category_code == 'fake'
-item._identity.category_entry_name == item.id.value
-item.id.unique_name.endswith('.fake.<id>.id')
-```
-
-Also test that a singleton-like item with `_category_code` and no
-`_category_entry_name` resolves category code but returns no entry name.
-
-### Current Category Parametrized Tests
-
-Add a parametrized test that instantiates representative loop item
-classes and verifies category code plus entry:
-
-```text
-Alias -> alias, label
-JointFitItem -> joint_fit, experiment_id
-SequentialFitExtractItem -> sequential_fit_extract, id
-AtomSite -> atom_site, label
-AtomSiteAniso -> atom_site_aniso, label
-LinkedPhase -> linked_phases, id
-LineSegment -> background, id
-PolynomialTerm -> background, id
-ExcludedRegion -> excluded_regions, id
-Refln -> refln, id
-PdCwlDataPoint -> pd_data, point_id
-PdTofDataPoint -> pd_data, point_id
-TotalDataPoint -> total_data, point_id
-```
-
-Do not instantiate heavy calculator-backed owner objects for this test;
-instantiate item classes directly.
-
-### Constraints Tests
-
-Update
-`tests/unit/easydiffraction/analysis/categories/test_constraints.py`.
-
-Required assertions:
-
-```python
-c = Constraint()
-c.expression = 'a = b + c'
-c.id = 'constraint_a'
-assert c.id.value == 'constraint_a'
-assert c.lhs_alias == 'a'
-assert c.rhs_expr == 'b + c'
-assert c._identity.category_entry_name == 'constraint_a'
-```
-
-Default id behavior:
-
-```python
-coll = Constraints()
-coll.create(expression='a = b + c')
-assert 'a' in coll.names
-assert coll['a'].id.value == 'a'
-assert coll['a'].rhs_expr == 'b + c'
-```
-
-Explicit id behavior:
-
-```python
-coll = Constraints()
-coll.create(id='c1', expression='a = b + c')
-assert 'c1' in coll.names
-assert coll['c1'].lhs_alias == 'a'
-```
-
-CIF serialization:
-
-```python
-cif = coll.as_cif
-assert '_constraint.id' in cif
-assert '_constraint.expression' in cif
-```
-
-Backward-compatible CIF loading:
-
-```python
-cif = '''
-data_analysis
-
-loop_
-_constraint.expression
-"a = b + c"
-'''
-```
-
-Load through the existing analysis/constraints loader and assert the
-loaded collection has key `a`.
-
-### Existing Round-Trip Tests
-
-Update tests that compare constraint CIF or constraint collection keys:
-
-- `tests/unit/easydiffraction/project/test_project_load.py`
-- `tests/unit/easydiffraction/io/cif/test_serialize_category_owner_baseline.py`
-- `tests/integration/fitting/test_project_load.py`
-
-Expected changes:
-
-- New saved analysis CIF includes `_constraint.id`.
-- Old expression-only fixtures, if any, still load.
-- Existing constraint workflows still work when no explicit id is
-  supplied.
-
-## Verification Commands
-
-Run the smallest useful checks first:
-
-```shell
-pixi run python -m pytest tests/unit/easydiffraction/analysis/categories/test_constraints.py
-pixi run python -m pytest tests/unit/easydiffraction/core/
-pixi run python -m pytest tests/unit/easydiffraction/io/cif/test_serialize_category_owner_baseline.py
-```
-
-Then run broader checks:
-
-```shell
-pixi run unit-tests
-pixi run integration-tests
-pixi run check
-```
-
-If the modified-file Prettier helper is still missing, format changed
-Markdown directly:
-
-```shell
-npx prettier --write --config=prettierrc.toml docs/dev/plans/loop-category-key-identity.md docs/dev/adrs/suggestions/loop-category-key-identity.md
-```
-
-## Local Availability Check
-
-The plan was checked against the current repository state:
-
-- `.github/copilot-instructions.md` exists.
-- `src/easydiffraction/core/category.py`,
-  `src/easydiffraction/core/identity.py`, and
-  `src/easydiffraction/io/cif/serialize.py` exist.
-- `tests/unit/easydiffraction/analysis/categories/test_constraints.py`
-  exists.
-- `tests/unit/easydiffraction/core/` exists.
-- `tests/unit/easydiffraction/io/cif/test_serialize_category_owner_baseline.py`
-  exists.
-- `pixi.toml` defines `fix`, `check`, `unit-tests`, `integration-tests`,
-  and `test-structure-check`.
-- `prettierrc.toml` and local Prettier are available.
-- `tools/nonpy_prettier_modified.py` is not present, so use the direct
-  `npx prettier --write --config=prettierrc.toml ...` fallback for
-  touched Markdown files.
-
-## Acceptance Criteria
-
-The implementation is done when all of these are true:
-
-- All current category codes are declared as `_category_code` on item
-  classes.
-- All current loop collection keys are declared as
-  `_category_entry_name` on item classes.
-- No current category item sets `self._identity.category_code` in its
-  constructor.
-- No current category item sets `self._identity.category_entry_name` in
-  its constructor.
-- `item._identity.category_entry_name` still works for all loop rows.
-- Descriptor `unique_name` values still include datablock, category,
-  entry, and descriptor name.
-- Constraints persist `_constraint.id`.
-- Constraints created without an explicit id still default to the left
-  hand alias.
-- Old constraints CIF without `_constraint.id` still loads and receives
-  deterministic ids.
-- Collection indexes are correct after CIF loading.
-
-## Suggested Pull Request
-
-Title:
-
-```text
-Use declarative category identity metadata
-```
-
-Description:
-
-```text
-This change makes category and loop-row identities easier to audit and
-keeps saved CIF identifiers explicit. Constraint rows gain a stable
-identifier while existing constraint expressions continue to work.
-```
diff --git a/docs/dev/plans/workspace-root-project-category.md b/docs/dev/plans/workspace-root-project-category.md
deleted file mode 100644
index 778fd7584..000000000
--- a/docs/dev/plans/workspace-root-project-category.md
+++ /dev/null
@@ -1,1229 +0,0 @@
-# Workspace Root and Project Category Migration Plan
-
-## Status
-
-Branch: `feature/workspace-root-project-category`
-
-ADR suggestion:
-
-```text
-docs/dev/adrs/suggestions/workspace-root-project-category.md
-```
-
-Two-phase workflow from `.github/copilot-instructions.md`:
-
-- Phase 1 - Implementation. Code and docs updates only. Do not create or
-  run tests unless the user explicitly asks.
-- Phase 2 - Verification. Add/update tests, then run the verification
-  commands listed near the end of this plan.
-
-Stop after Phase 1 and request review before starting Phase 2.
-
-Status checklist. Mark `[x]` only while implementing:
-
-```text
-Phase 1 - Implementation
-[ ] Phase 0: Confirm breaking-change approval.
-[ ] Phase 1: Rename root package and public facade to Workspace.
-[ ] Phase 2: Rename project-info access from info to project.
-[ ] Phase 3: Align project information fields with _project.* tags.
-[ ] Phase 4: Rename project config file to workspace.cif.
-[ ] Phase 5: Update root-object references across runtime code.
-[ ] Phase 6: Update docs, tutorials, and ADR references.
-[ ] Phase 7: Remove old public Project surface unless approved.
-[ ] Phase 1 review gate: present diff for approval.
-
-Phase 2 - Verification
-[ ] Move/update unit tests to workspace paths.
-[ ] Add workspace naming and CIF layout tests.
-[ ] pixi run test-structure-check
-[ ] pixi run fix
-[ ] pixi run check
-[ ] pixi run unit-tests
-[ ] pixi run integration-tests
-[ ] pixi run script-tests
-[ ] pixi run notebook-prepare
-[ ] pixi run notebook-tests
-```
-
-## Commit Discipline
-
-When an AI agent follows this plan, every completed Phase 1
-implementation step must be staged with explicit paths and committed
-locally before moving to the next implementation step or to the Phase 1
-review gate.
-
-Follow the **Commits** section of `.github/copilot-instructions.md`.
-
-Rules:
-
-- One commit per phase.
-- Keep each commit atomic and single-purpose.
-- Stage explicit paths only. Do not use `git add .`.
-- Use `git mv` for file and directory moves.
-- Do not stage unrelated user changes.
-- Do not stage generated artifacts unless the user explicitly asks.
-- If a serious uncovered design issue appears, stop and ask before
-  continuing.
-
-Suggested commit messages:
-
-```text
-Rename Project facade to Workspace
-Expose project information as workspace.project
-Align project metadata fields with CIF names
-Rename project config file to workspace.cif
-Update runtime references to Workspace
-Update docs for Workspace root API
-Remove old Project public API surface
-```
-
-## Goal
-
-Split the name "project" into two distinct concepts:
-
-1. `Workspace` - the top-level runtime facade, currently named
-   `Project`.
-2. `workspace.project` - the category that stores information about the
-   scientific project.
-
-Target public API:
-
-```python
-import easydiffraction as ed
-
-workspace = ed.Workspace(project_id='lbco_hrpt')
-workspace.project.id
-workspace.project.title = 'La0.5Ba0.5CoO3 at HRPT@PSI'
-workspace.rendering.table_engine = 'rich'
-workspace.verbosity = 'short'
-workspace.structures
-workspace.experiments
-workspace.analysis
-workspace.display
-workspace.summary
-workspace.save_as('lbco_hrpt')
-```
-
-Target workspace-level config file:
-
-```text
-workspace.cif
-```
-
-Target CIF tags inside `workspace.cif`:
-
-```cif
-_project.id
-_project.title
-_project.description
-_project.created
-_project.last_modified
-
-_rendering.chart_engine
-_rendering.table_engine
-
-_verbosity.level
-```
-
-Do not introduce `_meta.*` tags.
-
-Target saved layout:
-
-```text
-<workspace-dir>/
-|-- workspace.cif
-|-- structures/
-|   `-- cosio.cif
-|-- experiments/
-|   `-- d20.cif
-|-- analysis/
-|   `-- analysis.cif
-`-- summary/
-    `-- summary.cif
-```
-
-## Decisions Already Made
-
-Use these decisions unless the user explicitly changes the ADR before
-implementation:
-
-- The public root class becomes `Workspace`.
-- The public root import becomes
-  `from easydiffraction import Workspace`.
-- The public project-information category becomes `workspace.project`.
-- The public rendering category remains `workspace.rendering`.
-- The project-information category keeps semantic CIF tags `_project.*`.
-- The rendering category keeps semantic CIF tags `_rendering.*`.
-- The verbosity preference is serialized as `_verbosity.level`.
-- The saved singleton config file becomes `workspace.cif`.
-- The saved root is a workspace directory with a user-chosen filesystem
-  name; do not use `project` as the conceptual root name in new docs.
-- Do not use `project.cif`, `config.cif`, or `meta.cif` as the primary
-  singleton config file in the target layout.
-- The storage directory path belongs to `Workspace.path`, not
-  `workspace.project.path`.
-- The old `Project` public API is removed unless the user explicitly
-  approves an alias before implementation.
-- The category class name `ProjectInfo` is kept; only the public
-  attribute name (`info` → `project`) changes.
-- `ProjectInfo.path` is removed; the saved directory path lives on
-  `Workspace.path` only.
-- A first-class `Verbosity` category under `WorkspaceConfig` is required
-  (not optional) so that `_verbosity.level` round-trips through
-  `workspace.cif` like other singleton categories.
-- Source-tree imports may be temporarily inconsistent between phases
-  (for example, Phase 1 leaves `project_config_to_cif` imports until
-  Phase 4 renames the serializer functions). This is acceptable because
-  tests are not run in Phase 1. Each phase must still leave the source
-  importable at the end of the phase.
-
-## Current Shape
-
-The current implementation already has a category-owner based project
-configuration layer:
-
-```text
-src/easydiffraction/project/
-|-- project.py             # class Project
-|-- project_config.py      # class ProjectConfig
-|-- project_info.py        # ProjectInfo export
-|-- display.py             # class ProjectDisplay
-`-- categories/
-    |-- info/              # ProjectInfo category
-    `-- rendering/         # Rendering category
-```
-
-Current public API:
-
-```python
-project = ed.Project(name='my_project')
-project.info.title
-project.rendering.table_engine
-```
-
-Current saved config file:
-
-```text
-project.cif
-```
-
-## Target Shape
-
-Target implementation:
-
-```text
-src/easydiffraction/workspace/
-|-- workspace.py           # class Workspace
-|-- workspace_config.py    # class WorkspaceConfig
-|-- project_info.py        # ProjectInfo export
-|-- display.py             # class WorkspaceDisplay
-`-- categories/
-    |-- project/           # ProjectInfo category
-    |-- rendering/         # Rendering category
-    `-- verbosity/         # Verbosity category
-```
-
-Target public API:
-
-```python
-workspace = ed.Workspace(project_id='my_project')
-workspace.project.title
-workspace.rendering.table_engine
-workspace.verbosity = 'short'
-```
-
-## Out Of Scope
-
-Do not do these in this migration:
-
-- Do not add `_meta.*` CIF tags.
-- Do not use `project.cif`, `config.cif`, or `meta.cif` as the target
-  singleton settings file.
-- Do not redesign structure or experiment datablocks.
-- Do not change analysis fit-mode semantics.
-- Do not change calculator behavior.
-- Do not edit generated package-structure docs by hand.
-- Do not edit generated notebooks directly. Edit tutorial `.py` sources
-  and regenerate notebooks during Phase 2.
-- Do not keep a `Project = Workspace` compatibility alias unless the
-  user explicitly approves it.
-
-## Phase 0: Confirm Breaking-Change Approval
-
-This migration removes or replaces the public `Project` API unless the
-user approves a compatibility alias.
-
-Before changing code, ask the user to confirm:
-
-```text
-This migration removes ed.Project and replaces it with ed.Workspace.
-Should implementation proceed without a Project compatibility alias?
-```
-
-If the user asks for a compatibility alias, record that decision in this
-plan and in the ADR before implementation.
-
-Do not implement code before this approval gate.
-
-Commit: no commit required for this phase unless the plan or ADR is
-updated.
-
-## Phase 1: Rename Root Package And Public Facade
-
-### Objective
-
-Rename the top-level runtime facade and package from `project` to
-`workspace`.
-
-### Files Likely To Change
-
-- `src/easydiffraction/project/`
-- `src/easydiffraction/__init__.py`
-- `src/easydiffraction/__main__.py`
-- `src/easydiffraction/analysis/analysis.py`
-- `src/easydiffraction/analysis/sequential.py`
-- `src/easydiffraction/display/plotting.py`
-- `src/easydiffraction/summary/summary.py`
-- any source file importing `easydiffraction.project.*`
-
-### Steps
-
-1. Move the source package with `git mv` so history is preserved:
-
-   ```text
-   src/easydiffraction/project/
-   -> src/easydiffraction/workspace/
-   ```
-
-2. Rename files:
-
-   ```text
-   workspace/project.py
-   -> workspace/workspace.py
-
-   workspace/project_config.py
-   -> workspace/workspace_config.py
-   ```
-
-3. Rename classes:
-
-   ```text
-   Project -> Workspace
-   ProjectConfig -> WorkspaceConfig
-   ProjectDisplay -> WorkspaceDisplay
-   ```
-
-4. Rename class-level state inside the facade:
-
-   ```text
-   Project._current_project   -> Workspace._current_workspace
-   Project._loading           -> Workspace._loading      # name kept
-   Project.current_project_path() -> Workspace.current_workspace_path()
-   ```
-
-   Update the `ClassVar` annotation accordingly and update all internal
-   references (`type(self)._current_project = self`,
-   `cls._current_project`).
-
-5. Update the `varname()` fallback inside `__init__` so the default
-   variable name becomes `'workspace'` instead of `'project'`:
-
-   ```python
-   self._varname = 'workspace' if type(self)._loading else varname()
-   ```
-
-6. Update top-level import:
-
-   ```python
-   from easydiffraction.workspace.workspace import Workspace
-   ```
-
-7. Remove the old top-level `Project` import unless the user approved an
-   alias.
-
-8. Update type-checking imports:
-
-   ```python
-   from easydiffraction.workspace.workspace import Workspace
-   ```
-
-9. Update docstrings from "Project facade" to "Workspace facade" where
-   they describe the root object. Keep wording that talks about the
-   scientific project (titles, descriptions, identity) unchanged.
-
-10. Rename `io/ascii.py::extract_project_from_zip` to
-    `extract_workspace_from_zip` and update its re-exports in
-    `src/easydiffraction/io/__init__.py` and
-    `src/easydiffraction/__init__.py`. The function extracts a saved
-    workspace directory, not scientific project information.
-
-11. Note: `src/easydiffraction/io/cif/serialize.py` still defines
-    `project_config_to_cif`, `project_config_from_cif`, and
-    `project_to_cif` at this point. Leave those imports as-is in
-    `workspace.py`; they are renamed in Phase 4. The function
-    `project_info_to_cif` keeps its name.
-
-12. Run a source-only grep. Do not run tests in Phase 1:
-
-```shell
-rg -n "easydiffraction\\.project|\\bProject\\b|ProjectDisplay|ProjectConfig" src
-```
-
-For every match, decide whether it refers to:
-
-- the old root object, which should become `Workspace`;
-- the project-information category, which should remain project;
-- historical text that should be updated in docs later.
-
-### Stop Conditions
-
-Stop and ask if:
-
-- another public class named `Workspace` already exists;
-- package moves break imports in a way that would require compatibility
-  shims;
-- a file has both root-object `project` and category `project` meanings
-  that cannot be separated clearly.
-
-### Commit
-
-Stage explicit moved and edited files, then commit:
-
-```text
-Rename Project facade to Workspace
-```
-
-## Phase 2: Rename Project-Info Access From `info` To `project`
-
-### Objective
-
-Make the project-information category public as `workspace.project`
-instead of `workspace.info`.
-
-### Files Likely To Change
-
-- `src/easydiffraction/workspace/workspace_config.py`
-- `src/easydiffraction/workspace/workspace.py`
-- `src/easydiffraction/workspace/categories/info/`
-- `src/easydiffraction/workspace/project_info.py`
-- `src/easydiffraction/io/cif/serialize.py`
-- all code using `.info` for project information
-
-### Steps
-
-1. Rename category package:
-
-   ```text
-   src/easydiffraction/workspace/categories/info/
-   -> src/easydiffraction/workspace/categories/project/
-   ```
-
-2. Keep the category class name `ProjectInfo` for now. The class name is
-   explicit and avoids a confusing `Project` class after the root class
-   is renamed to `Workspace`.
-
-3. Rename imports:
-
-   ```python
-   from easydiffraction.workspace.categories.project import ProjectInfo
-   from easydiffraction.workspace.categories.project import ProjectInfoFactory
-   ```
-
-4. In `WorkspaceConfig`, rename:
-
-   ```text
-   _info -> _project
-   info -> project
-   ```
-
-5. In `Workspace`, rename:
-
-   ```text
-   _info -> _project
-   info -> project
-   ```
-
-6. Remove the public `.info` property unless the user approved a
-   compatibility alias.
-
-7. Update all runtime references:
-
-   ```text
-   workspace.info.title -> workspace.project.title
-   workspace.info.description -> workspace.project.description
-   workspace.info.update_last_modified() -> workspace.project.update_last_modified()
-   ```
-
-8. Run grep:
-
-   ```shell
-   rg -n "\\.info\\b|categories/info|categories\\.info" src
-   ```
-
-9. For every match, update it if it refers to project information. Leave
-   unrelated uses of the word "info" alone.
-
-### Stop Conditions
-
-Stop and ask if:
-
-- `info` appears as a different public concept unrelated to project
-  information;
-- removing `.info` would break a user-requested compatibility alias.
-
-### Commit
-
-```text
-Expose project information as workspace.project
-```
-
-## Phase 3: Align Project Information Fields With `_project.*`
-
-### Objective
-
-Expose project identity as `workspace.project.id`, matching CIF
-`_project.id`.
-
-Move the saved directory path to `workspace.path`, because it describes
-the workspace location and is not serialized project information.
-
-### Files Likely To Change
-
-- `src/easydiffraction/workspace/categories/project/default.py`
-- `src/easydiffraction/workspace/workspace.py`
-- `src/easydiffraction/io/cif/serialize.py`
-- `src/easydiffraction/summary/summary.py`
-- `src/easydiffraction/display/plotting.py`
-- any code using `.name` for project identity or `.project.path`
-
-### Steps
-
-1. In `ProjectInfo`, rename the public identity property and its setter:
-
-   ```text
-   name (getter)  -> id (getter)
-   name (setter)  -> id (setter)
-   ```
-
-   Do not also rename the internal descriptor attribute
-   `self._project_id`; it already matches the new public name.
-
-2. Keep the underlying CIF tag unchanged:
-
-   ```python
-   CifHandler(names=['_project.id'])
-   ```
-
-3. Update `ProjectInfo.unique_name` to return `self.id`.
-
-4. Keep `ProjectInfo._identity.category_code = 'project'` as-is.
-
-5. Update `project_info_to_cif()` and CIF loading helpers to use
-   `info.id`.
-
-6. Rename constructor arguments:
-
-   ```text
-   name -> project_id
-   ```
-
-   Apply this to:
-   - `Workspace.__init__`
-   - `WorkspaceConfig.__init__`
-   - `ProjectInfo.__init__`
-   - `ProjectInfoFactory.create(...)` call sites
-   - Default value: `'untitled_project'` (unchanged value, just the
-     parameter name changes)
-
-7. Add `Workspace.path` as the runtime storage path. Initialize
-   `self._path: pathlib.Path | None = None` in `Workspace.__init__`.
-
-   Suggested shape (match the surrounding `GuardedBase` pattern; do not
-   add `@typechecked` here because the setter accepts both `str` and
-   `pathlib.Path`):
-
-   ```python
-   @property
-   def path(self) -> pathlib.Path | None:
-       """Saved workspace directory."""
-       return self._path
-
-   @path.setter
-   def path(self, value: object) -> None:
-       self._path = pathlib.Path(value) if value is not None else None
-   ```
-
-8. Remove `ProjectInfo.path` (property, setter, and the `self._path`
-   attribute inside `ProjectInfo.__init__`) unless explicitly approved
-   as a compatibility alias.
-
-9. Update save/load logic across the codebase:
-
-   ```text
-   workspace.path
-   ```
-
-   should replace:
-
-   ```text
-   project.info.path        # old
-   workspace.project.path   # never used; do not introduce
-   ```
-
-   Concrete call sites include `Workspace.save`, `Workspace.load`,
-   `Workspace.current_workspace_path`, and any consumers in `analysis/`,
-   `display/`, `summary/`, and `io/`.
-
-10. Update messages and string representations:
-
-```text
-Workspace '<project_id>' (...)
-Saving workspace '<project_id>' to ...
-```
-
-11. Run grep:
-
-```shell
-rg -n "\\.name\\b|\\.path\\b|project_id|Project identifier" src/easydiffraction/workspace src/easydiffraction/io src/easydiffraction/display src/easydiffraction/summary
-```
-
-Inspect each match manually. Do not blindly replace every `.name`;
-structures and experiments still use `.name`.
-
-### Stop Conditions
-
-Stop and ask if:
-
-- a caller depends on `workspace.name` as a root-object property;
-- moving `path` out of `ProjectInfo` makes save/load unclear;
-- external saved fixtures require an approved compatibility path.
-
-### Commit
-
-```text
-Align project metadata fields with CIF names
-```
-
-## Phase 4: Rename Project Config File To `workspace.cif`
-
-### Objective
-
-Rename the saved singleton configuration file from `project.cif` to
-`workspace.cif`.
-
-### Files Likely To Change
-
-- `src/easydiffraction/workspace/workspace.py`
-- `src/easydiffraction/io/cif/serialize.py`
-- `src/easydiffraction/workspace/workspace_config.py`
-- `src/easydiffraction/workspace/categories/verbosity/`
-- CLI entry points in `src/easydiffraction/__main__.py`
-- docs that describe saved project directories
-- test fixtures in Phase 2
-
-### Steps
-
-1. Rename serializer functions in
-   `src/easydiffraction/io/cif/serialize.py` and every call site:
-
-   ```text
-   project_config_to_cif   -> workspace_config_to_cif
-   project_config_from_cif -> workspace_config_from_cif
-   project_to_cif          -> workspace_to_cif
-   ```
-
-   Call sites include `workspace.py` (formerly `project.py`),
-   `workspace_config.py`, and the serializer itself (the
-   `project_to_cif` body calls `project_config_to_cif`). After this step
-   the imports that were intentionally left stale in Phase 1 must
-   compile cleanly.
-
-   Do not rename `project_info_to_cif`; it serializes the `_project`
-   category and that name remains correct.
-
-2. Update `Workspace.save()` to write:
-
-   ```text
-   workspace.cif
-   ```
-
-3. Update `Workspace.load()` to read:
-
-   ```text
-   workspace.cif
-   ```
-
-4. Do not add `project.cif`, `config.cif`, or `meta.cif` fallbacks
-   unless the user approved a compatibility loader.
-
-5. Add a `Verbosity` category under
-   `src/easydiffraction/workspace/categories/verbosity/` following the
-   same shape as `rendering/` (a `default.py` with a `Verbosity`
-   `CategoryItem`, a `factory.py` with `VerbosityFactory`, and an
-   `__init__.py` that imports both to trigger registration).
-
-   The category owns one descriptor:
-
-   ```python
-   CifHandler(names=['_verbosity.level'])
-   ```
-
-   Bind it in `WorkspaceConfig.__init__` next to `_rendering`, and
-   expose it from `Workspace` so that the public access path remains:
-
-   ```python
-   workspace.verbosity = 'short'
-   workspace.verbosity              # -> 'short'
-   ```
-
-   The public `verbosity` getter/setter on `Workspace` reads and writes
-   the category's `level` descriptor (validated against `VerbosityEnum`)
-   and replaces the current `self._verbosity: VerbosityEnum`
-   runtime-only attribute. Remove that attribute.
-
-   Serialize it as:
-
-   ```cif
-   _verbosity.level short
-   ```
-
-6. Keep the contents semantic:
-
-   ```cif
-   _project.id
-   _project.title
-   _rendering.table_engine
-   _verbosity.level
-   ```
-
-7. Update logging and console output from `project.cif` to
-   `workspace.cif`.
-
-8. Run grep:
-
-   ```shell
-   rg -n "project\\.cif|config\\.cif|meta\\.cif|project_config_to_cif|project_config_from_cif|project_to_cif|verbosity" src docs tests
-   ```
-
-   Update source and docs only. Test files are handled in Phase 2 unless
-   the user explicitly asks otherwise.
-
-9. Saved on-disk fixtures under `data/` and `projects/` still contain
-   `project.cif` files (for example `data/lbco_project/project.cif`,
-   `projects/cosio/project.cif`). Do **not** edit or regenerate them in
-   Phase 1. They are inputs to integration/script tests and will either
-   be regenerated in Phase 2 or the relevant tests will be updated to
-   write fresh workspace directories. Flag any that block Phase 2 in the
-   review gate.
-
-### Stop Conditions
-
-Stop and ask if:
-
-- repository fixtures or tutorials contain saved directories that must
-  remain loadable without conversion;
-- the user wants a one-release compatibility loader.
-- the verbosity setting cannot be represented as a category without
-  weakening the public `workspace.verbosity` API.
-
-### Commit
-
-```text
-Rename project config file to workspace.cif
-```
-
-## Phase 5: Update Root-Object References Across Runtime Code
-
-### Objective
-
-Replace root-object variables and attributes named `project` with
-`workspace` where they refer to the top-level facade.
-
-Keep the word `project` where it refers to the project-information
-category or the scientific project itself.
-
-### Files Likely To Change
-
-- `src/easydiffraction/analysis/analysis.py`
-- `src/easydiffraction/analysis/sequential.py`
-- `src/easydiffraction/display/plotting.py`
-- `src/easydiffraction/project/display.py` after it has moved to
-  `workspace/display.py`
-- `src/easydiffraction/summary/summary.py`
-- `src/easydiffraction/__main__.py`
-- `src/easydiffraction/io/*`
-
-### Steps
-
-1. Rename root references in `Analysis`:
-
-   ```text
-   self.project -> self.workspace
-   analysis.project -> analysis.workspace
-   ```
-
-   This includes the constructor argument, the stored attribute, and any
-   public read-only property exposing the parent workspace.
-
-2. Rename display internals in `WorkspaceDisplay` (formerly
-   `ProjectDisplay`) and in any other class that stores a back-
-   reference to the root facade:
-
-   ```text
-   self._project -> self._workspace
-   _set_project(...) -> _set_workspace(...)
-   ```
-
-   Only do this when the stored object is the top-level runtime facade.
-   Do not touch `self._project_id` inside `ProjectInfo` or `_project.*`
-   CIF tags.
-
-3. Rename local variables in runtime code:
-
-   ```text
-   project = Workspace(...)
-   -> workspace = Workspace(...)
-   ```
-
-4. Keep scientific-project wording where appropriate:
-
-   ```text
-   workspace.project.title
-   project_id
-   _project.id
-   ```
-
-5. Update user-facing messages carefully. Good examples:
-
-   ```text
-   "Workspace directory not found"
-   "Saving workspace"
-   "Project title"
-   ```
-
-6. Run grep:
-
-   ```shell
-   rg -n "\\bproject\\b|\\bProject\\b|_project|ProjectDisplay" src/easydiffraction
-   ```
-
-7. Inspect each match. Do not replace `_project` CIF tags.
-
-### Stop Conditions
-
-Stop and ask if:
-
-- a name has both root-workspace and project-information meanings in the
-  same function and cannot be made clear;
-- renaming a method such as `_set_project` would require updating public
-  plugin or user code.
-
-### Commit
-
-```text
-Update runtime references to Workspace
-```
-
-## Phase 6: Update Docs, Tutorials, And ADR References
-
-### Objective
-
-Update user-facing and developer-facing documentation to describe the
-new root object and project-information category.
-
-### Files Likely To Change
-
-- `docs/dev/adrs/index.md`
-- `docs/dev/issues/open.md`
-- `docs/dev/adrs/accepted/*.md`
-- `docs/dev/adrs/suggestions/*.md`
-- `docs/docs/tutorials/*.py`
-- `README.md`
-- `CONTRIBUTING.md` only if it contains API examples
-
-Do not edit these by hand:
-
-- `docs/dev/package-structure/full.md`
-- `docs/dev/package-structure/short.md`
-- generated tutorial notebooks
-- generated `docs/site/` files
-
-### Steps
-
-1. Update the relevant accepted ADRs:
-
-   ```text
-   Project Facade and Persistence Layout
-   -> Workspace Facade and Persistence Layout
-   ```
-
-2. Update the affected ADR examples to use:
-
-   ```text
-   workspace.project      ProjectInfo
-   workspace.rendering    Rendering
-   workspace.verbosity    str
-   workspace.display      WorkspaceDisplay
-   ```
-
-3. Update saved layout examples:
-
-   ```text
-   workspace.cif
-   structures/
-   experiments/
-   analysis/
-   summary.cif
-   ```
-
-4. Update public examples:
-
-   ```python
-   workspace = ed.Workspace(project_id='lbco_hrpt')
-   workspace.project.title = '...'
-   ```
-
-5. Update ADRs that describe current API. Historical reasoning can keep
-   old names only if it is clearly historical and not presented as
-   current usage.
-
-6. Update tutorial `.py` files, not notebooks. Phase 2 will run
-   `pixi run notebook-prepare`.
-
-7. Run grep:
-
-   ```shell
-   rg -n "ed\\.Project|from easydiffraction import Project|project\\.info|project\\.rendering|ProjectDisplay|project\\.cif" docs README.md CONTRIBUTING.md
-   ```
-
-8. Inspect each match manually.
-
-### Stop Conditions
-
-Stop and ask if:
-
-- a tutorial title uses "Project" as ordinary English rather than API
-  naming;
-- historical ADRs would become misleading if edited mechanically.
-
-### Commit
-
-```text
-Update docs for Workspace root API
-```
-
-## Phase 7: Remove Old Public `Project` Surface Unless Approved
-
-### Objective
-
-Finish the breaking rename by removing old public imports and module
-paths unless the user approved compatibility.
-
-### Steps Without Compatibility Alias
-
-1. Ensure top-level `easydiffraction.__init__` exports `Workspace`, not
-   `Project`.
-
-2. Ensure no source imports from:
-
-   ```text
-   easydiffraction.project
-   ```
-
-3. Ensure no public source package remains at:
-
-   ```text
-   src/easydiffraction/project
-   ```
-
-4. Run grep:
-
-   ```shell
-   rg -n "from easydiffraction import Project|ed\\.Project|easydiffraction\\.project|\\bProject\\(" src docs tests tools README.md CONTRIBUTING.md
-   ```
-
-5. Any remaining match must be:
-   - historical text that intentionally names the old API; or
-   - a test that will be updated in Phase 2; or
-   - a generated artifact that should not be edited manually.
-
-### Steps With Approved Compatibility Alias
-
-Only do this if the user explicitly approved it.
-
-1. Add a temporary alias in `src/easydiffraction/__init__.py`:
-
-   ```python
-   Project = Workspace
-   ```
-
-2. Keep the alias undocumented unless the user asks for a migration
-   note.
-
-3. Add tests in Phase 2 proving both `Workspace` and `Project` construct
-   the same root object.
-
-### Commit
-
-Without alias:
-
-```text
-Remove old Project public API surface
-```
-
-With alias:
-
-```text
-Add Project alias for Workspace migration
-```
-
-## Phase 1 Review Gate
-
-After Phase 1 commits are complete:
-
-1. Run `git status --short`.
-2. Confirm only intended files are changed.
-3. Summarize:
-   - whether `Project` was removed or aliased;
-   - whether `workspace.cif` replaced `project.cif`;
-   - any files intentionally left for Phase 2 test updates;
-   - any unresolved questions.
-
-4. Stop and ask the user to review before starting Phase 2.
-
-Do not run the full verification suite until the user approves moving to
-Phase 2.
-
-## Phase 2: Verification And Tests
-
-Only start this phase after the user approves the Phase 1
-implementation.
-
-### Test Updates
-
-Move or update tests to mirror the new source tree:
-
-```text
-tests/unit/easydiffraction/project/
--> tests/unit/easydiffraction/workspace/
-```
-
-Update imports:
-
-```python
-from easydiffraction.workspace.workspace import Workspace
-from easydiffraction.workspace.display import WorkspaceDisplay
-```
-
-Update functional and integration tests:
-
-```python
-from easydiffraction import Workspace
-workspace = Workspace(project_id='...')
-```
-
-### New Tests To Add
-
-Add focused tests for:
-
-1. `from easydiffraction import Workspace`.
-2. `Workspace(project_id='p1').project.id == 'p1'`.
-3. `workspace.project.title` round-trips through `workspace.cif`.
-4. `workspace.rendering.table_engine` round-trips through
-   `workspace.cif`.
-5. `workspace.verbosity` round-trips through `workspace.cif`.
-6. `Workspace.save()` writes `workspace.cif`.
-7. `Workspace.load()` reads `workspace.cif`.
-8. `workspace.cif` contains `_project.id`, not `_meta.project_id`.
-9. `workspace.cif` contains `_rendering.table_engine`.
-10. `workspace.cif` contains `_verbosity.level`.
-11. `workspace.path` is set after `save_as()` and `load()`.
-12. `workspace.project` has no serialized path field.
-13. `project.cif`, `config.cif`, and `meta.cif` are not written unless
-    compatibility was approved.
-14. `ed.Project` is absent unless compatibility was approved.
-15. `Verbosity` category is registered via its factory and reachable
-    through `WorkspaceConfig` (parallel to `Rendering`).
-16. `ed.extract_workspace_from_zip` is importable and
-    `ed.extract_project_from_zip` is not (unless compatibility
-    approved).
-
-If compatibility alias was approved, add tests for:
-
-1. `from easydiffraction import Project`.
-2. `Project is Workspace` or equivalent behavior.
-3. Any approved `project.cif` fallback behavior.
-
-### Verification Commands
-
-Run in this order:
-
-```shell
-pixi run test-structure-check
-pixi run fix
-pixi run check
-pixi run unit-tests
-pixi run integration-tests
-pixi run script-tests
-pixi run notebook-prepare
-pixi run notebook-tests
-```
-
-If `pixi run fix` regenerates package-structure docs, accept those
-generated changes and do not hand-edit them.
-
-### Phase 2 Commit Suggestions
-
-Use one or more commits, depending on size:
-
-```text
-Update workspace unit tests
-Update tutorials for Workspace API
-Regenerate tutorial notebooks for Workspace API
-```
-
-## Grep Checklist
-
-Use this checklist before final review.
-
-Runtime root object should use `Workspace`:
-
-```shell
-rg -n "\\bProject\\b|ed\\.Project|from easydiffraction import Project" src tests docs tools README.md CONTRIBUTING.md
-```
-
-Project-information category should use `workspace.project`:
-
-```shell
-rg -n "\\.info\\b|workspace\\.project|project\\.info" src tests docs tools README.md CONTRIBUTING.md
-```
-
-CIF project category should stay `_project`:
-
-```shell
-rg -n "_meta\\.|_project\\." src tests docs tools README.md CONTRIBUTING.md
-```
-
-Saved config file should be `workspace.cif`:
-
-```shell
-rg -n "project\\.cif|config\\.cif|meta\\.cif|workspace\\.cif" src tests docs tools README.md CONTRIBUTING.md
-```
-
-Workspace verbosity should serialize as a workspace-level category:
-
-```shell
-rg -n "_verbosity|verbosity" src tests docs tools README.md CONTRIBUTING.md
-```
-
-Generated docs should not be manually edited:
-
-```shell
-git diff -- docs/site docs/dev/package-structure/full.md docs/dev/package-structure/short.md
-```
-
-If package-structure docs changed because of `pixi run fix`, that is
-expected. If `docs/site` changed, ask before staging.
-
-## Common Mistakes
-
-### Mistake: Renaming `_project.*` To `_workspace.*`
-
-Do not do this. The CIF category describes scientific project
-information, not the runtime facade.
-
-Correct:
-
-```cif
-_project.id
-_project.title
-```
-
-Incorrect:
-
-```cif
-_workspace.project_id
-_workspace.title
-```
-
-### Mistake: Introducing `_meta.*`
-
-Do not replace `_project.*` with `_meta.*`.
-
-Correct:
-
-```cif
-_project.title
-```
-
-Incorrect:
-
-```cif
-_meta.project_title
-```
-
-### Mistake: Keeping `project.cif` Or Switching To Generic File Names
-
-Do not use `project.cif`, `config.cif`, or `meta.cif` as the target
-singleton settings file. The file belongs to the workspace layer.
-
-Correct:
-
-```text
-workspace.cif
-```
-
-### Mistake: Blindly Replacing Every `project`
-
-Some uses of `project` should remain:
-
-- CIF tags such as `_project.id`
-- `workspace.project`
-- scientific project wording in prose
-- `ProjectInfo` class name, unless a later ADR changes it
-
-Only root-facade uses should become `workspace` or `Workspace`.
-
-### Mistake: Leaving Path On Project Information
-
-The saved directory path belongs to the workspace runtime state. It
-should be `workspace.path`, not `workspace.project.path`.
-
-### Mistake: Forgetting Facade Class-Level State
-
-When renaming `Project` to `Workspace`, the `ClassVar`
-`_current_project`, the `current_project_path()` classmethod, and the
-`varname()` fallback string `'project'` all live on the class itself and
-are easy to miss with a single search-and-replace. Rename them to
-`_current_workspace`, `current_workspace_path()`, and `'workspace'`
-respectively.
-
-### Mistake: Renaming `ProjectInfo._project_id`
-
-The internal descriptor attribute `self._project_id` inside
-`ProjectInfo` already matches the new public name `id` and stays as is.
-Only the public `name` property/setter becomes `id`.
-
-### Mistake: Editing Generated Notebooks Directly
-
-Tutorial notebooks are generated artifacts. Edit tutorial `.py` files,
-then run `pixi run notebook-prepare` in Phase 2.
-
-## Suggested Pull Request
-
-Title:
-
-```text
-Rename Project root object to Workspace
-```
-
-Description:
-
-```text
-This change separates the working EasyDiffraction workspace from the
-scientific project information stored inside it. Users now create a
-Workspace, while project title and description live under
-workspace.project and continue to serialize with clear _project.* CIF
-names.
-```
diff --git a/docs/docs/cli/index.md b/docs/docs/cli/index.md
index 5fc13c771..46c283879 100644
--- a/docs/docs/cli/index.md
+++ b/docs/docs/cli/index.md
@@ -48,6 +48,41 @@ List all available tutorial notebooks:
 python -m easydiffraction list-tutorials
 ```
 
+### List Example Data
+
+List all available example data files and downloadable project archives:
+
+```bash
+python -m easydiffraction list-data
+```
+
+The table includes the data ID, file name, record kind, and description.
+
+### Download Example Data
+
+Download a specific example data record by ID:
+
+```bash
+python -m easydiffraction download-data 3
+```
+
+For downloadable saved projects, the ZIP archive is extracted
+automatically, the ZIP file is removed, and the extracted project path
+is reported:
+
+```bash
+python -m easydiffraction download-data 30
+```
+
+This makes it possible to go straight from download to a project-first
+CLI command such as:
+
+```bash
+python -m easydiffraction EXTRACTED_PROJECT_DIR display
+```
+
+Use the extracted path printed by `download-data`.
+
 ### Download Tutorials
 
 Download a specific tutorial by ID:
@@ -66,12 +101,16 @@ Both commands accept `--destination` (`-d`) to specify the output
 directory (default: `tutorials/`) and `--overwrite` (`-o`) to replace
 existing files.
 
+`download-data` also accepts `--destination` (`-d`) and `--overwrite`
+(`-o`). For project archives, `--overwrite` replaces the extracted
+project directory before downloading and unpacking a fresh copy.
+
 ### Fit a Project
 
 Load a saved project and run structural refinement:
 
 ```bash
-python -m easydiffraction fit PROJECT_DIR
+python -m easydiffraction PROJECT_DIR fit
 ```
 
 `PROJECT_DIR` is the path to a project directory previously created by
@@ -86,5 +125,35 @@ Use the `--dry` flag to run the fit **without overwriting** the project
 files:
 
 ```bash
-python -m easydiffraction fit PROJECT_DIR --dry
+python -m easydiffraction PROJECT_DIR fit --dry
 ```
+
+EasyDiffraction also accepts the legacy subcommand-first form
+`python -m easydiffraction fit PROJECT_DIR`, but the project-first form
+is recommended because it makes it easy to rerun the same command and
+swap only the action.
+
+### Display a Project
+
+Load a saved project and show the outputs that match its current fit
+state and rendering backend:
+
+```bash
+python -m easydiffraction PROJECT_DIR display
+```
+
+For typical non-sequential projects this includes the latest fit
+results, parameter correlations, default pattern views, and when the
+saved state is Bayesian also posterior distributions and predictive
+checks. Plotly-only views such as posterior pair plots are shown only
+when the active chart engine is Plotly.
+
+### Undo the Last Fit
+
+The CLI already reserves the project-first undo command shape:
+
+```bash
+python -m easydiffraction PROJECT_DIR undo
+```
+
+This command currently reports that undo support is not implemented yet.
diff --git a/docs/docs/quick-reference/index.md b/docs/docs/quick-reference/index.md
index c035d3fe3..9c4d715bd 100644
--- a/docs/docs/quick-reference/index.md
+++ b/docs/docs/quick-reference/index.md
@@ -39,10 +39,15 @@ ed.show_version()
 Download a dataset by ID into a local directory:
 
 ```python
+ed.list_data()
+
 structure_path = ed.download_data(id=1, destination='data')
 data_path = ed.download_data(id=3, destination='data')
 ```
 
+Project archives are extracted automatically, and `download_data()`
+returns the extracted project directory path.
+
 For tutorial notebooks:
 
 ```python
@@ -398,8 +403,16 @@ project = ed.Project.load('lbco_hrpt')
 Run a saved project from the command line:
 
 ```bash
-python -m easydiffraction fit lbco_hrpt
-python -m easydiffraction fit lbco_hrpt --dry
+python -m easydiffraction lbco_hrpt fit
+python -m easydiffraction lbco_hrpt fit --dry
+python -m easydiffraction lbco_hrpt display
+```
+
+Load a saved example project straight from `download_data()`:
+
+```python
+saved_project_dir = ed.download_data(id=30, destination='projects')
+project = ed.Project.load(saved_project_dir)
 ```
 
 ## Command-Line Reminders
@@ -407,8 +420,11 @@ python -m easydiffraction fit lbco_hrpt --dry
 ```bash
 python -m easydiffraction --help
 python -m easydiffraction --version
+python -m easydiffraction list-data
+python -m easydiffraction download-data 30 --destination projects
 python -m easydiffraction list-tutorials
 python -m easydiffraction download-tutorial 1 --destination tutorials
 python -m easydiffraction download-all-tutorials --destination tutorials
-python -m easydiffraction fit PROJECT_DIR
+python -m easydiffraction PROJECT_DIR fit
+python -m easydiffraction PROJECT_DIR display
 ```
diff --git a/docs/docs/tutorials/ed-13.ipynb b/docs/docs/tutorials/ed-13.ipynb
index 616c536c5..ca636d82b 100644
--- a/docs/docs/tutorials/ed-13.ipynb
+++ b/docs/docs/tutorials/ed-13.ipynb
@@ -2657,7 +2657,7 @@
  ],
  "metadata": {
   "jupytext": {
-   "cell_metadata_filter": "tags,title,-all",
+   "cell_metadata_filter": "title,tags,-all",
    "main_language": "python",
    "notebook_metadata_filter": "-all"
   }
diff --git a/docs/docs/tutorials/ed-15.ipynb b/docs/docs/tutorials/ed-15.ipynb
index b6d82a319..e79bdd876 100644
--- a/docs/docs/tutorials/ed-15.ipynb
+++ b/docs/docs/tutorials/ed-15.ipynb
@@ -248,6 +248,17 @@
    "id": "23",
    "metadata": {},
    "outputs": [],
+   "source": [
+    "# Limit number of iterations to prevent long calculation time in this tutorial.\n",
+    "project.analysis.fitting.minimizer.max_iterations = 500"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "24",
+   "metadata": {},
+   "outputs": [],
    "source": [
     "# Start refinement. All parameters, which have standard uncertainties\n",
     "# in the input CIF files, are refined by default.\n",
@@ -257,7 +268,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "24",
+   "id": "25",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -268,7 +279,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "25",
+   "id": "26",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -278,7 +289,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "26",
+   "id": "27",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -288,7 +299,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "27",
+   "id": "28",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -297,7 +308,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "28",
+   "id": "29",
    "metadata": {},
    "source": [
     "## Step 5: Perform Analysis (ADP aniso)"
@@ -306,7 +317,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "29",
+   "id": "30",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -317,7 +328,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "30",
+   "id": "31",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -330,7 +341,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "31",
+   "id": "32",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -340,7 +351,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "32",
+   "id": "33",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -350,7 +361,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "33",
+   "id": "34",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -360,7 +371,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "34",
+   "id": "35",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -370,7 +381,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "35",
+   "id": "36",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -380,7 +391,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "36",
+   "id": "37",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -390,7 +401,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "37",
+   "id": "38",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/docs/tutorials/ed-15.py b/docs/docs/tutorials/ed-15.py
index 7359b712a..03dc35fa7 100644
--- a/docs/docs/tutorials/ed-15.py
+++ b/docs/docs/tutorials/ed-15.py
@@ -79,6 +79,10 @@
 # %%
 project.analysis.fitting.minimizer_type = 'bumps'
 
+# %%
+# Limit number of iterations to prevent long calculation time in this tutorial.
+project.analysis.fitting.minimizer.max_iterations = 500
+
 # %%
 # Start refinement. All parameters, which have standard uncertainties
 # in the input CIF files, are refined by default.
diff --git a/docs/docs/tutorials/ed-17.ipynb b/docs/docs/tutorials/ed-17.ipynb
index f32b0371f..e471d99a7 100644
--- a/docs/docs/tutorials/ed-17.ipynb
+++ b/docs/docs/tutorials/ed-17.ipynb
@@ -58,7 +58,8 @@
    "source": [
     "## Step 1: Define Project\n",
     "\n",
-    "The project object manages structures, experiments, and analysis."
+    "The project object manages structures, experiments, analysis, display,\n",
+    "and other related components."
    ]
   },
   {
@@ -68,7 +69,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "project = ed.Project()"
+    "project = ed.Project(name='cosio_d20')\n",
+    "analysis = project.analysis\n",
+    "display = project.display"
    ]
   },
   {
@@ -87,7 +90,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.save_as('projects/cosio', temporary=False)"
+    "project.save_as(dir_path='projects/cosio_d20')"
    ]
   },
   {
@@ -111,7 +114,7 @@
    "outputs": [],
    "source": [
     "project.structures.create(name='cosio')\n",
-    "structure = project.structures['cosio']"
+    "struct = project.structures['cosio']"
    ]
   },
   {
@@ -129,8 +132,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "structure.space_group.name_h_m = 'P n m a'\n",
-    "structure.space_group.it_coordinate_system_code = 'abc'"
+    "struct.space_group.name_h_m = 'P n m a'\n",
+    "struct.space_group.it_coordinate_system_code = 'abc'"
    ]
   },
   {
@@ -148,9 +151,9 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "structure.cell.length_a = 10.31\n",
-    "structure.cell.length_b = 6.0\n",
-    "structure.cell.length_c = 4.79"
+    "struct.cell.length_a = 10.31\n",
+    "struct.cell.length_b = 6.0\n",
+    "struct.cell.length_c = 4.79"
    ]
   },
   {
@@ -168,7 +171,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "structure.atom_sites.create(\n",
+    "struct.atom_sites.create(\n",
     "    label='Co1',\n",
     "    type_symbol='Co',\n",
     "    fract_x=0,\n",
@@ -177,7 +180,7 @@
     "    wyckoff_letter='a',\n",
     "    adp_iso=0.3,\n",
     ")\n",
-    "structure.atom_sites.create(\n",
+    "struct.atom_sites.create(\n",
     "    label='Co2',\n",
     "    type_symbol='Co',\n",
     "    fract_x=0.279,\n",
@@ -186,7 +189,7 @@
     "    wyckoff_letter='c',\n",
     "    adp_iso=0.3,\n",
     ")\n",
-    "structure.atom_sites.create(\n",
+    "struct.atom_sites.create(\n",
     "    label='Si',\n",
     "    type_symbol='Si',\n",
     "    fract_x=0.094,\n",
@@ -195,7 +198,7 @@
     "    wyckoff_letter='c',\n",
     "    adp_iso=0.34,\n",
     ")\n",
-    "structure.atom_sites.create(\n",
+    "struct.atom_sites.create(\n",
     "    label='O1',\n",
     "    type_symbol='O',\n",
     "    fract_x=0.091,\n",
@@ -204,7 +207,7 @@
     "    wyckoff_letter='c',\n",
     "    adp_iso=0.63,\n",
     ")\n",
-    "structure.atom_sites.create(\n",
+    "struct.atom_sites.create(\n",
     "    label='O2',\n",
     "    type_symbol='O',\n",
     "    fract_x=0.448,\n",
@@ -213,7 +216,7 @@
     "    wyckoff_letter='c',\n",
     "    adp_iso=0.59,\n",
     ")\n",
-    "structure.atom_sites.create(\n",
+    "struct.atom_sites.create(\n",
     "    label='O3',\n",
     "    type_symbol='O',\n",
     "    fract_x=0.164,\n",
@@ -246,7 +249,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "zip_path = ed.download_data(id=27, destination='data')"
+    "zip_path = ed.download_data(id=25, destination='data')"
    ]
   },
   {
@@ -427,28 +430,28 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "structure.cell.length_a.free = True\n",
-    "structure.cell.length_b.free = True\n",
-    "structure.cell.length_c.free = True\n",
+    "struct.cell.length_a.free = True\n",
+    "struct.cell.length_b.free = True\n",
+    "struct.cell.length_c.free = True\n",
     "\n",
-    "structure.atom_sites['Co2'].fract_x.free = True\n",
-    "structure.atom_sites['Co2'].fract_z.free = True\n",
-    "structure.atom_sites['Si'].fract_x.free = True\n",
-    "structure.atom_sites['Si'].fract_z.free = True\n",
-    "structure.atom_sites['O1'].fract_x.free = True\n",
-    "structure.atom_sites['O1'].fract_z.free = True\n",
-    "structure.atom_sites['O2'].fract_x.free = True\n",
-    "structure.atom_sites['O2'].fract_z.free = True\n",
-    "structure.atom_sites['O3'].fract_x.free = True\n",
-    "structure.atom_sites['O3'].fract_y.free = True\n",
-    "structure.atom_sites['O3'].fract_z.free = True\n",
+    "struct.atom_sites['Co2'].fract_x.free = True\n",
+    "struct.atom_sites['Co2'].fract_z.free = True\n",
+    "struct.atom_sites['Si'].fract_x.free = True\n",
+    "struct.atom_sites['Si'].fract_z.free = True\n",
+    "struct.atom_sites['O1'].fract_x.free = True\n",
+    "struct.atom_sites['O1'].fract_z.free = True\n",
+    "struct.atom_sites['O2'].fract_x.free = True\n",
+    "struct.atom_sites['O2'].fract_z.free = True\n",
+    "struct.atom_sites['O3'].fract_x.free = True\n",
+    "struct.atom_sites['O3'].fract_y.free = True\n",
+    "struct.atom_sites['O3'].fract_z.free = True\n",
     "\n",
-    "structure.atom_sites['Co1'].adp_iso.free = True\n",
-    "structure.atom_sites['Co2'].adp_iso.free = True\n",
-    "structure.atom_sites['Si'].adp_iso.free = True\n",
-    "structure.atom_sites['O1'].adp_iso.free = True\n",
-    "structure.atom_sites['O2'].adp_iso.free = True\n",
-    "structure.atom_sites['O3'].adp_iso.free = True"
+    "struct.atom_sites['Co1'].adp_iso.free = True\n",
+    "struct.atom_sites['Co2'].adp_iso.free = True\n",
+    "struct.atom_sites['Si'].adp_iso.free = True\n",
+    "struct.atom_sites['O1'].adp_iso.free = True\n",
+    "struct.atom_sites['O2'].adp_iso.free = True\n",
+    "struct.atom_sites['O3'].adp_iso.free = True"
    ]
   },
   {
@@ -488,13 +491,13 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.aliases.create(\n",
+    "analysis.aliases.create(\n",
     "    label='biso_Co1',\n",
-    "    param=structure.atom_sites['Co1'].adp_iso,\n",
+    "    param=struct.atom_sites['Co1'].adp_iso,\n",
     ")\n",
-    "project.analysis.aliases.create(\n",
+    "analysis.aliases.create(\n",
     "    label='biso_Co2',\n",
-    "    param=structure.atom_sites['Co2'].adp_iso,\n",
+    "    param=struct.atom_sites['Co2'].adp_iso,\n",
     ")"
    ]
   },
@@ -513,7 +516,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.constraints.create(expression='biso_Co2 = biso_Co1')"
+    "analysis.constraints.create(expression='biso_Co2 = biso_Co1')"
    ]
   },
   {
@@ -531,7 +534,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.fitting.minimizer_type = 'bumps (lm)'"
+    "analysis.fitting.minimizer_type = 'bumps (lm)'"
    ]
   },
   {
@@ -554,13 +557,23 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.fit()"
+    "analysis.fit()"
    ]
   },
   {
-   "cell_type": "markdown",
+   "cell_type": "code",
+   "execution_count": null,
    "id": "44",
    "metadata": {},
+   "outputs": [],
+   "source": [
+    "display.fit.results()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "45",
+   "metadata": {},
    "source": [
     "#### Show parameter correlations"
    ]
@@ -568,16 +581,16 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "45",
+   "id": "46",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.display.fit.correlations()"
+    "display.fit.correlations()"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "46",
+   "id": "47",
    "metadata": {},
    "source": [
     "#### Compare measured and calculated patterns for the first fit."
@@ -586,16 +599,16 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "47",
+   "id": "48",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.display.pattern(expt_name='d20')"
+    "display.pattern(expt_name='d20')"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "48",
+   "id": "49",
    "metadata": {},
    "source": [
     "#### Run Sequential Fitting\n",
@@ -607,7 +620,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "49",
+   "id": "50",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -616,7 +629,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "50",
+   "id": "51",
    "metadata": {
     "lines_to_next_cell": 2
    },
@@ -629,7 +642,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "51",
+   "id": "52",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -639,11 +652,11 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52",
+   "id": "53",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.sequential_fit_extract.create(\n",
+    "analysis.sequential_fit_extract.create(\n",
     "    id='temperature',\n",
     "    target=temperature,\n",
     "    pattern=r'^TEMP\\s+([0-9.]+)',\n",
@@ -653,7 +666,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "53",
+   "id": "54",
    "metadata": {},
    "source": [
     "Set the sequential fitting parameters."
@@ -662,19 +675,19 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "54",
+   "id": "55",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.fitting_mode_type = 'sequential'\n",
-    "project.analysis.sequential_fit.data_dir = scan_data_dir\n",
-    "project.analysis.sequential_fit.max_workers = 'auto'\n",
-    "project.analysis.sequential_fit.reverse = True"
+    "analysis.fitting_mode_type = 'sequential'\n",
+    "analysis.sequential_fit.data_dir = scan_data_dir\n",
+    "analysis.sequential_fit.max_workers = 'auto'\n",
+    "analysis.sequential_fit.reverse = True"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "55",
+   "id": "56",
    "metadata": {},
    "source": [
     "Run the sequential fit over all data files in the scan directory."
@@ -683,16 +696,16 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "56",
+   "id": "57",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.fit()"
+    "analysis.fit()"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "57",
+   "id": "58",
    "metadata": {},
    "source": [
     "#### Replay a Dataset\n",
@@ -703,17 +716,17 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "58",
+   "id": "59",
    "metadata": {},
    "outputs": [],
    "source": [
     "project.apply_params_from_csv(row_index=0)\n",
-    "project.display.pattern(expt_name='d20')"
+    "display.pattern(expt_name='d20')"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "59",
+   "id": "60",
    "metadata": {},
    "source": [
     "\n",
@@ -723,17 +736,17 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "60",
+   "id": "61",
    "metadata": {},
    "outputs": [],
    "source": [
     "project.apply_params_from_csv(row_index=-1)\n",
-    "project.display.pattern(expt_name='d20')"
+    "display.pattern(expt_name='d20')"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "61",
+   "id": "62",
    "metadata": {},
    "source": [
     "#### Plot Parameter Evolution\n",
@@ -743,7 +756,27 @@
   },
   {
    "cell_type": "markdown",
-   "id": "62",
+   "id": "63",
+   "metadata": {},
+   "source": [
+    "Plot fit quality metrics vs. temperature."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "64",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "display.fit.series(analysis.fit_result.success, versus=temperature)\n",
+    "display.fit.series(analysis.fit_result.reduced_chi_square, versus=temperature)\n",
+    "display.fit.series(analysis.fit_result.iterations, versus=temperature)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "65",
    "metadata": {},
    "source": [
     "Plot unit cell parameters vs. temperature."
@@ -752,18 +785,18 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "63",
+   "id": "66",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.display.fit.series(structure.cell.length_a, versus=temperature)\n",
-    "project.display.fit.series(structure.cell.length_b, versus=temperature)\n",
-    "project.display.fit.series(structure.cell.length_c, versus=temperature)"
+    "display.fit.series(struct.cell.length_a, versus=temperature)\n",
+    "display.fit.series(struct.cell.length_b, versus=temperature)\n",
+    "display.fit.series(struct.cell.length_c, versus=temperature)"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "64",
+   "id": "67",
    "metadata": {},
    "source": [
     "Plot isotropic displacement parameters vs. temperature."
@@ -772,35 +805,20 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "65",
+   "id": "68",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['Co1'].adp_iso,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['Si'].adp_iso,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['O1'].adp_iso,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['O2'].adp_iso,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['O3'].adp_iso,\n",
-    "    versus=temperature,\n",
-    ")"
+    "display.fit.series(struct.atom_sites['Co1'].adp_iso, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['Si'].adp_iso, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['O1'].adp_iso, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['O2'].adp_iso, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['O3'].adp_iso, versus=temperature)"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "66",
+   "id": "69",
    "metadata": {},
    "source": [
     "Plot selected fractional coordinates vs. temperature."
@@ -809,30 +827,15 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "67",
+   "id": "70",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['Co2'].fract_x,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['Co2'].fract_z,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['O1'].fract_z,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['O2'].fract_z,\n",
-    "    versus=temperature,\n",
-    ")\n",
-    "project.display.fit.series(\n",
-    "    structure.atom_sites['O3'].fract_z,\n",
-    "    versus=temperature,\n",
-    ")"
+    "display.fit.series(struct.atom_sites['Co2'].fract_x, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['Co2'].fract_z, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['O1'].fract_z, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['O2'].fract_z, versus=temperature)\n",
+    "display.fit.series(struct.atom_sites['O3'].fract_z, versus=temperature)"
    ]
   }
  ],
diff --git a/docs/docs/tutorials/ed-17.py b/docs/docs/tutorials/ed-17.py
index 24e648c56..93f691fb7 100644
--- a/docs/docs/tutorials/ed-17.py
+++ b/docs/docs/tutorials/ed-17.py
@@ -16,17 +16,20 @@
 # %% [markdown]
 # ## Step 1: Define Project
 #
-# The project object manages structures, experiments, and analysis.
+# The project object manages structures, experiments, analysis, display,
+# and other related components.
 
 # %%
-project = ed.Project()
+project = ed.Project(name='cosio_d20')
+analysis = project.analysis
+display = project.display
 
 # %% [markdown]
 # The project must be saved before running sequential fitting, so that
 # results can be written to `analysis/results.csv`.
 
 # %%
-project.save_as(dir_path='projects/cosio', temporary=False)
+project.save_as(dir_path='projects/cosio_d20')
 
 # %% [markdown]
 # ## Step 2: Define Crystal Structure
@@ -38,28 +41,28 @@
 
 # %%
 project.structures.create(name='cosio')
-structure = project.structures['cosio']
+struct = project.structures['cosio']
 
 # %% [markdown]
 # #### Set Space Group
 
 # %%
-structure.space_group.name_h_m = 'P n m a'
-structure.space_group.it_coordinate_system_code = 'abc'
+struct.space_group.name_h_m = 'P n m a'
+struct.space_group.it_coordinate_system_code = 'abc'
 
 # %% [markdown]
 # #### Set Unit Cell
 
 # %%
-structure.cell.length_a = 10.31
-structure.cell.length_b = 6.0
-structure.cell.length_c = 4.79
+struct.cell.length_a = 10.31
+struct.cell.length_b = 6.0
+struct.cell.length_c = 4.79
 
 # %% [markdown]
 # #### Set Atom Sites
 
 # %%
-structure.atom_sites.create(
+struct.atom_sites.create(
     label='Co1',
     type_symbol='Co',
     fract_x=0,
@@ -68,7 +71,7 @@
     wyckoff_letter='a',
     adp_iso=0.3,
 )
-structure.atom_sites.create(
+struct.atom_sites.create(
     label='Co2',
     type_symbol='Co',
     fract_x=0.279,
@@ -77,7 +80,7 @@
     wyckoff_letter='c',
     adp_iso=0.3,
 )
-structure.atom_sites.create(
+struct.atom_sites.create(
     label='Si',
     type_symbol='Si',
     fract_x=0.094,
@@ -86,7 +89,7 @@
     wyckoff_letter='c',
     adp_iso=0.34,
 )
-structure.atom_sites.create(
+struct.atom_sites.create(
     label='O1',
     type_symbol='O',
     fract_x=0.091,
@@ -95,7 +98,7 @@
     wyckoff_letter='c',
     adp_iso=0.63,
 )
-structure.atom_sites.create(
+struct.atom_sites.create(
     label='O2',
     type_symbol='O',
     fract_x=0.448,
@@ -104,7 +107,7 @@
     wyckoff_letter='c',
     adp_iso=0.59,
 )
-structure.atom_sites.create(
+struct.atom_sites.create(
     label='O3',
     type_symbol='O',
     fract_x=0.164,
@@ -125,7 +128,7 @@
 # #### Download Measured Data
 
 # %%
-zip_path = ed.download_data(id=27, destination='data')
+zip_path = ed.download_data(id=25, destination='data')
 
 # %% [markdown]
 # #### Extract Data Files
@@ -205,28 +208,28 @@
 # #### Set Free Parameters
 
 # %%
-structure.cell.length_a.free = True
-structure.cell.length_b.free = True
-structure.cell.length_c.free = True
+struct.cell.length_a.free = True
+struct.cell.length_b.free = True
+struct.cell.length_c.free = True
 
-structure.atom_sites['Co2'].fract_x.free = True
-structure.atom_sites['Co2'].fract_z.free = True
-structure.atom_sites['Si'].fract_x.free = True
-structure.atom_sites['Si'].fract_z.free = True
-structure.atom_sites['O1'].fract_x.free = True
-structure.atom_sites['O1'].fract_z.free = True
-structure.atom_sites['O2'].fract_x.free = True
-structure.atom_sites['O2'].fract_z.free = True
-structure.atom_sites['O3'].fract_x.free = True
-structure.atom_sites['O3'].fract_y.free = True
-structure.atom_sites['O3'].fract_z.free = True
+struct.atom_sites['Co2'].fract_x.free = True
+struct.atom_sites['Co2'].fract_z.free = True
+struct.atom_sites['Si'].fract_x.free = True
+struct.atom_sites['Si'].fract_z.free = True
+struct.atom_sites['O1'].fract_x.free = True
+struct.atom_sites['O1'].fract_z.free = True
+struct.atom_sites['O2'].fract_x.free = True
+struct.atom_sites['O2'].fract_z.free = True
+struct.atom_sites['O3'].fract_x.free = True
+struct.atom_sites['O3'].fract_y.free = True
+struct.atom_sites['O3'].fract_z.free = True
 
-structure.atom_sites['Co1'].adp_iso.free = True
-structure.atom_sites['Co2'].adp_iso.free = True
-structure.atom_sites['Si'].adp_iso.free = True
-structure.atom_sites['O1'].adp_iso.free = True
-structure.atom_sites['O2'].adp_iso.free = True
-structure.atom_sites['O3'].adp_iso.free = True
+struct.atom_sites['Co1'].adp_iso.free = True
+struct.atom_sites['Co2'].adp_iso.free = True
+struct.atom_sites['Si'].adp_iso.free = True
+struct.atom_sites['O1'].adp_iso.free = True
+struct.atom_sites['O2'].adp_iso.free = True
+struct.atom_sites['O3'].adp_iso.free = True
 
 # %%
 expt.linked_phases['cosio'].scale.free = True
@@ -247,26 +250,26 @@
 # Set aliases for parameters.
 
 # %%
-project.analysis.aliases.create(
+analysis.aliases.create(
     label='biso_Co1',
-    param=structure.atom_sites['Co1'].adp_iso,
+    param=struct.atom_sites['Co1'].adp_iso,
 )
-project.analysis.aliases.create(
+analysis.aliases.create(
     label='biso_Co2',
-    param=structure.atom_sites['Co2'].adp_iso,
+    param=struct.atom_sites['Co2'].adp_iso,
 )
 
 # %% [markdown]
 # Set constraints.
 
 # %%
-project.analysis.constraints.create(expression='biso_Co2 = biso_Co1')
+analysis.constraints.create(expression='biso_Co2 = biso_Co1')
 
 # %% [markdown]
 # #### Set Minimizer
 
 # %%
-project.analysis.fitting.minimizer_type = 'bumps (lm)'
+analysis.fitting.minimizer_type = 'bumps (lm)'
 
 # %% [markdown]
 # #### Run Single Fitting
@@ -277,19 +280,22 @@
 # if the initial parameters are far from optimal.
 
 # %%
-project.analysis.fit()
+analysis.fit()
+
+# %%
+display.fit.results()
 
 # %% [markdown]
 # #### Show parameter correlations
 
 # %%
-project.display.fit.correlations()
+display.fit.correlations()
 
 # %% [markdown]
 # #### Compare measured and calculated patterns for the first fit.
 
 # %%
-project.display.pattern(expt_name='d20')
+display.pattern(expt_name='d20')
 
 # %% [markdown]
 # #### Run Sequential Fitting
@@ -310,7 +316,7 @@
 temperature = 'diffrn.ambient_temperature'
 
 # %%
-project.analysis.sequential_fit_extract.create(
+analysis.sequential_fit_extract.create(
     id='temperature',
     target=temperature,
     pattern=r'^TEMP\s+([0-9.]+)',
@@ -321,16 +327,16 @@
 # Set the sequential fitting parameters.
 
 # %%
-project.analysis.fitting_mode_type = 'sequential'
-project.analysis.sequential_fit.data_dir = scan_data_dir
-project.analysis.sequential_fit.max_workers = 'auto'
-project.analysis.sequential_fit.reverse = True
+analysis.fitting_mode_type = 'sequential'
+analysis.sequential_fit.data_dir = scan_data_dir
+analysis.sequential_fit.max_workers = 'auto'
+analysis.sequential_fit.reverse = True
 
 # %% [markdown]
 # Run the sequential fit over all data files in the scan directory.
 
 # %%
-project.analysis.fit()
+analysis.fit()
 
 # %% [markdown]
 # #### Replay a Dataset
@@ -339,7 +345,7 @@
 
 # %%
 project.apply_params_from_csv(row_index=0)
-project.display.pattern(expt_name='d20')
+display.pattern(expt_name='d20')
 
 # %% [markdown]
 #
@@ -347,67 +353,45 @@
 
 # %%
 project.apply_params_from_csv(row_index=-1)
-project.display.pattern(expt_name='d20')
+display.pattern(expt_name='d20')
 
 # %% [markdown]
 # #### Plot Parameter Evolution
 #
 # Reuse the extracted diffrn path as the x-axis in the following plots.
 
+# %% [markdown]
+# Plot fit quality metrics vs. temperature.
+
+# %%
+display.fit.series(analysis.fit_result.success, versus=temperature)
+display.fit.series(analysis.fit_result.reduced_chi_square, versus=temperature)
+display.fit.series(analysis.fit_result.iterations, versus=temperature)
+
 # %% [markdown]
 # Plot unit cell parameters vs. temperature.
 
 # %%
-project.display.fit.series(structure.cell.length_a, versus=temperature)
-project.display.fit.series(structure.cell.length_b, versus=temperature)
-project.display.fit.series(structure.cell.length_c, versus=temperature)
+display.fit.series(struct.cell.length_a, versus=temperature)
+display.fit.series(struct.cell.length_b, versus=temperature)
+display.fit.series(struct.cell.length_c, versus=temperature)
 
 # %% [markdown]
 # Plot isotropic displacement parameters vs. temperature.
 
 # %%
-project.display.fit.series(
-    structure.atom_sites['Co1'].adp_iso,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['Si'].adp_iso,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['O1'].adp_iso,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['O2'].adp_iso,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['O3'].adp_iso,
-    versus=temperature,
-)
+display.fit.series(struct.atom_sites['Co1'].adp_iso, versus=temperature)
+display.fit.series(struct.atom_sites['Si'].adp_iso, versus=temperature)
+display.fit.series(struct.atom_sites['O1'].adp_iso, versus=temperature)
+display.fit.series(struct.atom_sites['O2'].adp_iso, versus=temperature)
+display.fit.series(struct.atom_sites['O3'].adp_iso, versus=temperature)
 
 # %% [markdown]
 # Plot selected fractional coordinates vs. temperature.
 
 # %%
-project.display.fit.series(
-    structure.atom_sites['Co2'].fract_x,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['Co2'].fract_z,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['O1'].fract_z,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['O2'].fract_z,
-    versus=temperature,
-)
-project.display.fit.series(
-    structure.atom_sites['O3'].fract_z,
-    versus=temperature,
-)
+display.fit.series(struct.atom_sites['Co2'].fract_x, versus=temperature)
+display.fit.series(struct.atom_sites['Co2'].fract_z, versus=temperature)
+display.fit.series(struct.atom_sites['O1'].fract_z, versus=temperature)
+display.fit.series(struct.atom_sites['O2'].fract_z, versus=temperature)
+display.fit.series(struct.atom_sites['O3'].fract_z, versus=temperature)
diff --git a/docs/docs/tutorials/ed-18.ipynb b/docs/docs/tutorials/ed-18.ipynb
index c1200010a..7473bc6e5 100644
--- a/docs/docs/tutorials/ed-18.ipynb
+++ b/docs/docs/tutorials/ed-18.ipynb
@@ -53,8 +53,7 @@
    "outputs": [],
    "source": [
     "from easydiffraction import Project\n",
-    "from easydiffraction import download_data\n",
-    "from easydiffraction import extract_project_from_zip"
+    "from easydiffraction import download_data"
    ]
   },
   {
@@ -62,7 +61,7 @@
    "id": "4",
    "metadata": {},
    "source": [
-    "## Download Project Archive"
+    "## Download Saved Project"
    ]
   },
   {
@@ -72,31 +71,13 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "zip_path = download_data(id=30, destination='data')"
+    "project_dir = download_data(id=36, destination='projects')"
    ]
   },
   {
    "cell_type": "markdown",
    "id": "6",
    "metadata": {},
-   "source": [
-    "## Extract Project"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "7",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "project_dir = extract_project_from_zip(zip_path, destination='data')"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "id": "8",
-   "metadata": {},
    "source": [
     "## Load Project"
    ]
@@ -104,7 +85,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9",
+   "id": "7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +94,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "10",
+   "id": "8",
    "metadata": {},
    "source": [
     "## Perform Analysis"
@@ -122,7 +103,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "11",
+   "id": "9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -131,7 +112,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "12",
+   "id": "10",
    "metadata": {},
    "source": [
     "## Show Results"
@@ -140,7 +121,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13",
+   "id": "11",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -150,7 +131,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "14",
+   "id": "12",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -160,7 +141,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "15",
+   "id": "13",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/docs/tutorials/ed-18.py b/docs/docs/tutorials/ed-18.py
index 7c4fff9fc..31e903c04 100644
--- a/docs/docs/tutorials/ed-18.py
+++ b/docs/docs/tutorials/ed-18.py
@@ -17,19 +17,12 @@
 # %%
 from easydiffraction import Project
 from easydiffraction import download_data
-from easydiffraction import extract_project_from_zip
 
 # %% [markdown]
-# ## Download Project Archive
+# ## Download Saved Project
 
 # %%
-zip_path = download_data(id=30, destination='data')
-
-# %% [markdown]
-# ## Extract Project
-
-# %%
-project_dir = extract_project_from_zip(zip_path, destination='data')
+project_dir = download_data(id=36, destination='projects')
 
 # %% [markdown]
 # ## Load Project
diff --git a/docs/docs/tutorials/ed-2.ipynb b/docs/docs/tutorials/ed-2.ipynb
index e3338eec8..1212cbcc7 100644
--- a/docs/docs/tutorials/ed-2.ipynb
+++ b/docs/docs/tutorials/ed-2.ipynb
@@ -519,6 +519,24 @@
    "source": [
     "project.display.pattern(expt_name='hrpt')"
    ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "42",
+   "metadata": {},
+   "source": [
+    "## Step 7: Save Project"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "43",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.save_as('projects/lbco_hrpt')"
+   ]
   }
  ],
  "metadata": {
diff --git a/docs/docs/tutorials/ed-2.py b/docs/docs/tutorials/ed-2.py
index a8e9e3d22..220ebae9c 100644
--- a/docs/docs/tutorials/ed-2.py
+++ b/docs/docs/tutorials/ed-2.py
@@ -227,3 +227,9 @@
 
 # %%
 project.display.pattern(expt_name='hrpt')
+
+# %% [markdown]
+# ## Step 7: Save Project
+
+# %%
+project.save_as('projects/lbco_hrpt')
diff --git a/docs/docs/tutorials/ed-20.ipynb b/docs/docs/tutorials/ed-20.ipynb
index 688fc593a..086d7d185 100644
--- a/docs/docs/tutorials/ed-20.ipynb
+++ b/docs/docs/tutorials/ed-20.ipynb
@@ -459,7 +459,7 @@
    "outputs": [],
    "source": [
     "project = Project(name='beer')\n",
-    "project.save_as(dir_path='beer_mcstas')"
+    "project.save_as(dir_path='projects/beer_mcstas')"
    ]
   },
   {
diff --git a/docs/docs/tutorials/ed-21.ipynb b/docs/docs/tutorials/ed-21.ipynb
index 4905c6f09..689c3d57a 100644
--- a/docs/docs/tutorials/ed-21.ipynb
+++ b/docs/docs/tutorials/ed-21.ipynb
@@ -74,7 +74,10 @@
     "\n",
     "The project object keeps structures, experiments, fit settings, and\n",
     "plotting utilities together in a single place. We will build the full\n",
-    "workflow inside this object."
+    "workflow inside this object.\n",
+    "\n",
+    "Save the project to a directory early on so that you can easily reload\n",
+    "it later if needed."
    ]
   },
   {
@@ -88,9 +91,19 @@
    ]
   },
   {
-   "cell_type": "markdown",
+   "cell_type": "code",
+   "execution_count": null,
    "id": "6",
    "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.save_as('projects/lbco_hrpt_bayesian')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "7",
+   "metadata": {},
    "source": [
     "## Step 2: Build the Structural Model\n",
     "\n",
@@ -102,7 +115,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "7",
+   "id": "8",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -112,7 +125,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "8",
+   "id": "9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -122,7 +135,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9",
+   "id": "10",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -133,7 +146,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "10",
+   "id": "11",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -142,7 +155,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "11",
+   "id": "12",
    "metadata": {},
    "source": [
     "The atom-site definitions below form the starting structural model. The\n",
@@ -153,7 +166,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "12",
+   "id": "13",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -203,7 +216,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "13",
+   "id": "14",
    "metadata": {},
    "source": [
     "## Step 3: Define the Diffraction Experiment\n",
@@ -215,7 +228,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "14",
+   "id": "15",
    "metadata": {},
    "source": [
     "Download the measured data from the repository. Alternatively, you\n",
@@ -226,7 +239,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "15",
+   "id": "16",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -235,7 +248,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "16",
+   "id": "17",
    "metadata": {},
    "source": [
     "Create the experiment object and specify the sample form, beam mode,\n",
@@ -245,7 +258,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "17",
+   "id": "18",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -261,7 +274,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "18",
+   "id": "19",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -270,7 +283,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "19",
+   "id": "20",
    "metadata": {},
    "source": [
     "Link the structural phase to the experiment."
@@ -279,7 +292,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "20",
+   "id": "21",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -288,7 +301,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "21",
+   "id": "22",
    "metadata": {},
    "source": [
     "Set instrument and peak profile parameters.\n",
@@ -300,7 +313,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "22",
+   "id": "23",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -311,7 +324,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "23",
+   "id": "24",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -323,7 +336,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "24",
+   "id": "25",
    "metadata": {},
    "source": [
     "Add background points and excluded regions.\n",
@@ -335,7 +348,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "25",
+   "id": "26",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -348,7 +361,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "26",
+   "id": "27",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -358,7 +371,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "27",
+   "id": "28",
    "metadata": {},
    "source": [
     "## Step 4: Run an Initial Local Refinement\n",
@@ -378,7 +391,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "28",
+   "id": "29",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -388,7 +401,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "29",
+   "id": "30",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -400,7 +413,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "30",
+   "id": "31",
    "metadata": {},
    "source": [
     "We choose the BUMPS Levenberg-Marquardt minimizer as a fast local\n",
@@ -411,7 +424,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "31",
+   "id": "32",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -421,7 +434,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "32",
+   "id": "33",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -431,7 +444,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "33",
+   "id": "34",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -441,7 +454,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "34",
+   "id": "35",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -450,7 +463,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "35",
+   "id": "36",
    "metadata": {},
    "source": [
     "The correlation plot shows how strongly the fitted parameters move\n",
@@ -462,7 +475,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "36",
+   "id": "37",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -472,7 +485,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "37",
+   "id": "38",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -481,7 +494,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "38",
+   "id": "39",
    "metadata": {},
    "source": [
     "## Step 5: Prepare for Bayesian Sampling\n",
@@ -503,7 +516,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "39",
+   "id": "40",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -512,7 +525,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "40",
+   "id": "41",
    "metadata": {},
    "source": [
     "Set fit bounds for all free parameters using the default multiplier of\n",
@@ -524,7 +537,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "41",
+   "id": "42",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -534,7 +547,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "42",
+   "id": "43",
    "metadata": {},
    "source": [
     "Displaying the free parameters again is a convenient way to confirm\n",
@@ -545,7 +558,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "43",
+   "id": "44",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -554,7 +567,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "44",
+   "id": "45",
    "metadata": {},
    "source": [
     "## Step 6: Configure and Run DREAM\n",
@@ -580,7 +593,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "45",
+   "id": "46",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -590,7 +603,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "46",
+   "id": "47",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -600,17 +613,18 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "47",
+   "id": "48",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.fitting.minimizer.steps = 300  # lower than the default 3000"
+    "project.analysis.fitting.minimizer.steps = 100  # lower than the default 3000\n",
+    "project.analysis.fitting.minimizer.burn = 20  # lower than the default 600"
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "48",
+   "id": "49",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -619,7 +633,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "49",
+   "id": "50",
    "metadata": {},
    "source": [
     "## Step 7: Inspect Bayesian Results\n",
@@ -632,7 +646,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "50",
+   "id": "51",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -641,7 +655,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "51",
+   "id": "52",
    "metadata": {},
    "source": [
     "The correlation and posterior-pair plots are complementary:\n",
@@ -658,7 +672,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "52",
+   "id": "53",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -668,7 +682,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "53",
+   "id": "54",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -677,7 +691,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "54",
+   "id": "55",
    "metadata": {},
    "source": [
     "The one-dimensional posterior distributions below make it easier to\n",
@@ -688,7 +702,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "55",
+   "id": "56",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -697,7 +711,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "56",
+   "id": "57",
    "metadata": {},
    "source": [
     "Finally, the posterior predictive plot propagates the sampled parameter\n",
@@ -709,7 +723,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "57",
+   "id": "58",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -718,7 +732,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "58",
+   "id": "59",
    "metadata": {},
    "source": [
     "A final zoomed measured-vs-calculated plot is useful for checking how\n",
@@ -729,7 +743,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "59",
+   "id": "60",
    "metadata": {},
    "outputs": [],
    "source": [
diff --git a/docs/docs/tutorials/ed-21.py b/docs/docs/tutorials/ed-21.py
index 87f003e46..df32bcabe 100644
--- a/docs/docs/tutorials/ed-21.py
+++ b/docs/docs/tutorials/ed-21.py
@@ -33,10 +33,16 @@
 # The project object keeps structures, experiments, fit settings, and
 # plotting utilities together in a single place. We will build the full
 # workflow inside this object.
+#
+# Save the project to a directory early on so that you can easily reload
+# it later if needed.
 
 # %%
 project = ed.Project()
 
+# %%
+project.save_as('projects/lbco_hrpt_bayesian')
+
 # %% [markdown]
 # ## Step 2: Build the Structural Model
 #
@@ -291,7 +297,8 @@
 project.analysis.fitting.minimizer_type = 'bumps (dream)'
 
 # %%
-project.analysis.fitting.minimizer.steps = 300  # lower than the default 3000
+project.analysis.fitting.minimizer.steps = 100  # lower than the default 3000
+project.analysis.fitting.minimizer.burn = 20  # lower than the default 600
 
 # %%
 project.analysis.fit()
diff --git a/docs/docs/tutorials/ed-22.ipynb b/docs/docs/tutorials/ed-22.ipynb
index bacb7c777..a8d511132 100644
--- a/docs/docs/tutorials/ed-22.ipynb
+++ b/docs/docs/tutorials/ed-22.ipynb
@@ -482,7 +482,8 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.analysis.fitting.minimizer.steps = 500  # lower than the default 3000"
+    "project.analysis.fitting.minimizer.steps = 100  # lower than the default 3000\n",
+    "project.analysis.fitting.minimizer.burn = 20  # lower than the default 600"
    ]
   },
   {
diff --git a/docs/docs/tutorials/ed-22.py b/docs/docs/tutorials/ed-22.py
index da43c6d52..cbc08f74e 100644
--- a/docs/docs/tutorials/ed-22.py
+++ b/docs/docs/tutorials/ed-22.py
@@ -219,7 +219,8 @@
 project.analysis.fitting.minimizer_type = 'bumps (dream)'
 
 # %%
-project.analysis.fitting.minimizer.steps = 500  # lower than the default 3000
+project.analysis.fitting.minimizer.steps = 100  # lower than the default 3000
+project.analysis.fitting.minimizer.burn = 20  # lower than the default 600
 
 # %%
 project.analysis.fit()
diff --git a/docs/docs/tutorials/ed-23.ipynb b/docs/docs/tutorials/ed-23.ipynb
index d20045c6c..11a28c791 100644
--- a/docs/docs/tutorials/ed-23.ipynb
+++ b/docs/docs/tutorials/ed-23.ipynb
@@ -56,10 +56,11 @@
    "id": "4",
    "metadata": {},
    "source": [
-    "## Download Saved Project Archive\n",
+    "## Download Saved Project\n",
     "\n",
-    "The archive should contain a saved project directory with a partially\n",
-    "completed sequential fit, including `analysis/results.csv`."
+    "The returned path points directly to the saved project directory with\n",
+    "a partially completed sequential fit, including\n",
+    "`analysis/results.csv`."
    ]
   },
   {
@@ -69,34 +70,13 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "zip_path = ed.download_data(id=34, destination='data')"
+    "project_dir = ed.download_data(id=37, destination='projects')"
    ]
   },
   {
    "cell_type": "markdown",
    "id": "6",
    "metadata": {},
-   "source": [
-    "## Extract Project\n",
-    "\n",
-    "Extract the saved project directory locally. For a project you\n",
-    "already have on disk, set `project_dir` directly instead."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "id": "7",
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "project_dir = ed.extract_project_from_zip(zip_path, destination='projects')"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "id": "8",
-   "metadata": {},
    "source": [
     "## Load Saved Project"
    ]
@@ -104,7 +84,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "9",
+   "id": "7",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -113,7 +93,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "10",
+   "id": "8",
    "metadata": {},
    "source": [
     "## Resume Sequential Analysis\n",
@@ -127,7 +107,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "11",
+   "id": "9",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -136,7 +116,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "12",
+   "id": "10",
    "metadata": {},
    "source": [
     "## Replay Fitted Datasets\n",
@@ -147,7 +127,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "13",
+   "id": "11",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -157,7 +137,7 @@
   },
   {
    "cell_type": "markdown",
-   "id": "14",
+   "id": "12",
    "metadata": {},
    "source": [
     "\n",
@@ -167,7 +147,7 @@
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "15",
+   "id": "13",
    "metadata": {},
    "outputs": [],
    "source": [
@@ -177,55 +157,70 @@
   },
   {
    "cell_type": "markdown",
-   "id": "16",
+   "id": "14",
    "metadata": {},
    "source": [
     "## Plot Parameter Evolution\n",
     "\n",
     "Use the same persisted diffrn path stored in `analysis/results.csv`\n",
-    "for the x-axis. Omitting `param` plots every fitted parameter one\n",
-    "after another."
+    "for the x-axis."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "17",
+   "id": "15",
    "metadata": {},
    "outputs": [],
    "source": [
     "temperature = 'diffrn.ambient_temperature'"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "16",
+   "metadata": {},
+   "source": [
+    "Plot fit quality metrics vs. temperature."
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "18",
+   "id": "17",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.display.fit.series(versus=temperature)"
+    "project.display.fit.series(\n",
+    "    project.analysis.fit_result.success,\n",
+    "    versus=temperature,\n",
+    ")\n",
+    "project.display.fit.series(\n",
+    "    project.analysis.fit_result.reduced_chi_square,\n",
+    "    versus=temperature,\n",
+    ")\n",
+    "project.display.fit.series(\n",
+    "    project.analysis.fit_result.iterations,\n",
+    "    versus=temperature,\n",
+    ")"
    ]
   },
   {
    "cell_type": "markdown",
-   "id": "19",
+   "id": "18",
    "metadata": {},
    "source": [
-    "## Save Project\n",
-    "\n",
-    "Save the updated project so the appended `analysis/results.csv` and\n",
-    "refreshed summary files remain on disk."
+    "Omitting `param` plots every fitted parameter one after another."
    ]
   },
   {
    "cell_type": "code",
    "execution_count": null,
-   "id": "20",
+   "id": "19",
    "metadata": {},
    "outputs": [],
    "source": [
-    "project.save()"
+    "project.display.fit.series(versus=temperature)"
    ]
   }
  ],
diff --git a/docs/docs/tutorials/ed-23.py b/docs/docs/tutorials/ed-23.py
index 659cbcbb1..a4dbc8ed7 100644
--- a/docs/docs/tutorials/ed-23.py
+++ b/docs/docs/tutorials/ed-23.py
@@ -14,22 +14,14 @@
 import easydiffraction as ed
 
 # %% [markdown]
-# ## Download Saved Project Archive
+# ## Download Saved Project
 #
-# The archive should contain a saved project directory with a partially
-# completed sequential fit, including `analysis/results.csv`.
+# The returned path points directly to the saved project directory with
+# a partially completed sequential fit, including
+# `analysis/results.csv`.
 
 # %%
-zip_path = ed.download_data(id=34, destination='data')
-
-# %% [markdown]
-# ## Extract Project
-#
-# Extract the saved project directory locally. For a project you
-# already have on disk, set `project_dir` directly instead.
-
-# %%
-project_dir = ed.extract_project_from_zip(zip_path, destination='projects')
+project_dir = ed.download_data(id=37, destination='projects')
 
 # %% [markdown]
 # ## Load Saved Project
@@ -69,11 +61,30 @@
 # ## Plot Parameter Evolution
 #
 # Use the same persisted diffrn path stored in `analysis/results.csv`
-# for the x-axis. Omitting `param` plots every fitted parameter one
-# after another.
+# for the x-axis.
 
 # %%
 temperature = 'diffrn.ambient_temperature'
 
+# %% [markdown]
+# Plot fit quality metrics vs. temperature.
+
+# %%
+project.display.fit.series(
+    project.analysis.fit_result.success,
+    versus=temperature,
+)
+project.display.fit.series(
+    project.analysis.fit_result.reduced_chi_square,
+    versus=temperature,
+)
+project.display.fit.series(
+    project.analysis.fit_result.iterations,
+    versus=temperature,
+)
+
+# %% [markdown]
+# Omitting `param` plots every fitted parameter one after another.
+
 # %%
 project.display.fit.series(versus=temperature)
diff --git a/docs/docs/tutorials/ed-24.ipynb b/docs/docs/tutorials/ed-24.ipynb
new file mode 100644
index 000000000..745d962f8
--- /dev/null
+++ b/docs/docs/tutorials/ed-24.ipynb
@@ -0,0 +1,223 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "0",
+   "metadata": {
+    "tags": [
+     "hide-in-docs"
+    ]
+   },
+   "outputs": [],
+   "source": [
+    "# Check whether easydiffraction is installed; install it if needed.\n",
+    "# Required for remote environments such as Google Colab.\n",
+    "import importlib.util\n",
+    "\n",
+    "if importlib.util.find_spec('easydiffraction') is None:\n",
+    "    %pip install easydiffraction"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "1",
+   "metadata": {},
+   "source": [
+    "# Load Saved Bayesian Project: LBCO, HRPT\n",
+    "\n",
+    "This tutorial shows how to reopen the Bayesian project created in\n",
+    "`ed-21.py` and inspect the saved fit results without rerunning DREAM.\n",
+    "\n",
+    "The project already contains posterior samples together with cached\n",
+    "posterior density, pair, and predictive data, so the plots below are\n",
+    "restored directly from disk."
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "2",
+   "metadata": {},
+   "source": [
+    "## Import Library"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "3",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import easydiffraction as ed"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "4",
+   "metadata": {},
+   "source": [
+    "## Download Saved Project\n",
+    "\n",
+    "The returned path points directly to the saved project directory with\n",
+    "the completed Bayesian fit and persisted posterior samples and plot\n",
+    "caches."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "5",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project_dir = ed.download_data(id=35, destination='projects')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "6",
+   "metadata": {},
+   "source": [
+    "## Load the Saved Bayesian Project\n",
+    "\n",
+    "Loading restores the persisted fit state, posterior samples, and plot\n",
+    "caches. No new fit is launched in this tutorial."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "7",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project = ed.Project.load(project_dir)"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "8",
+   "metadata": {},
+   "source": [
+    "## Review the Saved Fit Summary\n",
+    "\n",
+    "The fit summary reports the committed point estimate, sampler\n",
+    "settings, convergence diagnostics, and posterior parameter summaries\n",
+    "from the saved Bayesian run."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "9",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.display.fit.results()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "10",
+   "metadata": {},
+   "source": [
+    "## Show Correlations\n",
+    "\n",
+    "The correlation matrix is restored from the saved project state."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "11",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.display.fit.correlations()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "12",
+   "metadata": {},
+   "source": [
+    "## Inspect Posterior Densities and Pair Structure\n",
+    "\n",
+    "The pair plot and one-dimensional posterior distributions now load\n",
+    "from the persisted caches generated when the Bayesian fit was saved."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "13",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.display.posterior.pairs()"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "14",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.display.posterior.distribution()"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "15",
+   "metadata": {},
+   "source": [
+    "## Plot Posterior Predictive Checks\n",
+    "\n",
+    "The posterior predictive view reuses the cached predictive summary\n",
+    "stored in the project rather than recalculating it on first display.\n",
+    "It overlays the 95% credible interval propagated from the posterior\n",
+    "samples."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "16",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.display.posterior.predictive(expt_name='hrpt')"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "id": "17",
+   "metadata": {},
+   "source": [
+    "A zoomed view is useful for checking the propagated uncertainty in a\n",
+    "narrow region of the diffraction pattern."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "id": "18",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "project.display.posterior.predictive(expt_name='hrpt', x_min=92, x_max=93)"
+   ]
+  }
+ ],
+ "metadata": {
+  "jupytext": {
+   "cell_metadata_filter": "-all",
+   "main_language": "python",
+   "notebook_metadata_filter": "-all"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}
diff --git a/docs/docs/tutorials/ed-24.py b/docs/docs/tutorials/ed-24.py
new file mode 100644
index 000000000..9e5d40f87
--- /dev/null
+++ b/docs/docs/tutorials/ed-24.py
@@ -0,0 +1,82 @@
+# %% [markdown]
+# # Load Saved Bayesian Project: LBCO, HRPT
+#
+# This tutorial shows how to reopen the Bayesian project created in
+# `ed-21.py` and inspect the saved fit results without rerunning DREAM.
+#
+# The project already contains posterior samples together with cached
+# posterior density, pair, and predictive data, so the plots below are
+# restored directly from disk.
+
+# %% [markdown]
+# ## Import Library
+
+# %%
+import easydiffraction as ed
+
+# %% [markdown]
+# ## Download Saved Project
+#
+# The returned path points directly to the saved project directory with
+# the completed Bayesian fit and persisted posterior samples and plot
+# caches.
+
+# %%
+project_dir = ed.download_data(id=35, destination='projects')
+
+# %% [markdown]
+# ## Load the Saved Bayesian Project
+#
+# Loading restores the persisted fit state, posterior samples, and plot
+# caches. No new fit is launched in this tutorial.
+
+# %%
+project = ed.Project.load(project_dir)
+
+# %% [markdown]
+# ## Review the Saved Fit Summary
+#
+# The fit summary reports the committed point estimate, sampler
+# settings, convergence diagnostics, and posterior parameter summaries
+# from the saved Bayesian run.
+
+# %%
+project.display.fit.results()
+
+# %% [markdown]
+# ## Show Correlations
+#
+# The correlation matrix is restored from the saved project state.
+
+# %%
+project.display.fit.correlations()
+
+# %% [markdown]
+# ## Inspect Posterior Densities and Pair Structure
+#
+# The pair plot and one-dimensional posterior distributions now load
+# from the persisted caches generated when the Bayesian fit was saved.
+
+# %%
+project.display.posterior.pairs()
+
+# %%
+project.display.posterior.distribution()
+
+# %% [markdown]
+# ## Plot Posterior Predictive Checks
+#
+# The posterior predictive view reuses the cached predictive summary
+# stored in the project rather than recalculating it on first display.
+# It overlays the 95% credible interval propagated from the posterior
+# samples.
+
+# %%
+project.display.posterior.predictive(expt_name='hrpt')
+
+# %% [markdown]
+# A zoomed view is useful for checking the propagated uncertainty in a
+# narrow region of the diffraction pattern.
+
+# %%
+project.display.posterior.predictive(expt_name='hrpt', x_min=92, x_max=93)
diff --git a/docs/docs/tutorials/index.md b/docs/docs/tutorials/index.md
index a89601095..b789645d0 100644
--- a/docs/docs/tutorials/index.md
+++ b/docs/docs/tutorials/index.md
@@ -17,10 +17,6 @@ The tutorials are organized into the following categories:
 
 ## Getting Started
 
-- [LBCO `quick` `load`](ed-18.ipynb) – The most minimal example showing
-  how to load a previously saved project from a directory and run
-  refinement. Useful when a project has already been set up and saved in
-  a prior session.
 - [LBCO `quick` `code`](ed-2.ipynb) – A minimal example intended as a
   quick reference for users already familiar with the EasyDiffraction
   API or who want to see an example refinement when both the structure
@@ -42,6 +38,19 @@ The tutorials are organized into the following categories:
   descriptions of every step, making it suitable for users who are new
   to EasyDiffraction or those who prefer a more guided approach.
 
+## Load Project
+
+- [LBCO Single Fit](ed-18.ipynb) – The most minimal example showing how
+  to load a previously saved project from a directory and continue
+  working with it.
+- [Co2SiO4 Sequential Fit](ed-23.ipynb) – Resumes a sequential
+  refinement from an existing `analysis/results.csv` after an incomplete
+  previous run.
+- [LBCO Bayesian Display](ed-24.ipynb) – Shows how to load the saved
+  project after a Bayesian analysis and inspect the persisted fit
+  summary, correlation matrix, posterior plots, and predictive checks
+  without rerunning MCMC sampling.
+
 ## Powder Diffraction
 
 - [Co2SiO4 `pd-neut-cwl`](ed-5.ipynb) – Demonstrates a Rietveld
@@ -89,9 +98,6 @@ The tutorials are organized into the following categories:
 - [Co2SiO4 Temperature scan](ed-17.ipynb) – Sequential Rietveld
   refinement of Co2SiO4 using constant wavelength neutron powder
   diffraction data from D20 at ILL across a temperature scan.
-- [Co2SiO4 Temperature scan, resumed](ed-23.ipynb) – Continue a saved
-  sequential refinement of Co2SiO4 from an existing
-  `analysis/results.csv` after an incomplete previous run.
 
 ## Simulated Data
 
diff --git a/docs/docs/user-guide/analysis-workflow/project.md b/docs/docs/user-guide/analysis-workflow/project.md
index 5e8e8b417..72f77dc18 100644
--- a/docs/docs/user-guide/analysis-workflow/project.md
+++ b/docs/docs/user-guide/analysis-workflow/project.md
@@ -14,7 +14,7 @@ contribution from multiple **structures**.
 EasyDiffraction allows you to:
 
 - **Manually create** a new project by specifying its metadata.
-- **Load an existing project** from a file (**CIF** format).
+- **Load an existing saved project** from its project directory.
 
 Below are instructions on how to set up a project in EasyDiffraction. It
 is assumed that you have already imported the `easydiffraction` package,
@@ -55,14 +55,15 @@ you can just call the `save`:
 project.save()
 ```
 
-## Loading a Project from CIF
+## Loading a Saved Project
 
-If you have an existing project, you can load it directly from a CIF
-file. This is useful for reusing previously defined projects or sharing
-them with others.
+If you have an existing saved project, load it from the project
+directory created by `project.save_as()` or `project.save()`. This is
+useful for continuing a previous session or reusing a downloaded saved
+project.
 
 ```python
-project.load('data/lbco_hrpt.cif')
+project = ed.Project.load('lbco_hrpt')
 ```
 
 ## Project Structure
@@ -82,9 +83,10 @@ The example below illustrates a typical **project structure** for a
 ├── 📁 experiments    - Folder with experiment settings and measured data.
 │   ├── 📄 <span class="orange"><b>hrpt.cif</b></span>   - Instrumental parameters, calculator selection and measured data from HRPT@PSI.
 │   └── ...
-├── 📄 <span class="orange"><b>analysis.cif</b></span>   - Settings for data analysis (minimizer, fit mode, etc.).
-└── 📁 summary
-    └── 📄 report.cif - Summary report after structure refinement.
+├── 📁 analysis       - Analysis settings and optional persisted Bayesian arrays.
+│   ├── 📄 <span class="orange"><b>analysis.cif</b></span> - Settings for data analysis (minimizer, fit mode, constraints, persisted fit state).
+│   └── 📄 <span class="orange"><b>results.h5</b></span>   - Optional Bayesian sidecar with posterior and predictive arrays.
+└── 📄 <span class="orange"><b>summary.cif</b></span>    - Summary report after structure refinement.
 </pre>
 </div>
 
@@ -110,13 +112,16 @@ This file stores project-level metadata and display configuration.
 
 <div class="cif">
 <pre>
-data_<span class="red"><b>La0.5Ba0.5CoO3</b></span>
-
+<span class="blue"><b>_project</b>.id</span>          lbco_hrpt
 <span class="blue"><b>_project</b>.title</span>       "La0.5Ba0.5CoO3 from neutron diffraction at HRPT@PSI"
 <span class="blue"><b>_project</b>.description</span> "neutrons, powder, constant wavelength, HRPT@PSI"
 
-<span class="blue"><b>_display</b>.chart_engine</span>  asciichartpy
-<span class="blue"><b>_display</b>.table_engine</span>   rich
+<span class="blue"><b>_project</b>.created</span>     "18 May 2026 10:15:00"
+<span class="blue"><b>_project</b>.last_modified</span> "18 May 2026 10:20:00"
+
+<span class="blue"><b>_rendering</b>.chart_engine</span> auto
+<span class="blue"><b>_rendering</b>.table_engine</span> auto
+<span class="blue"><b>_verbosity</b>.fit</span>         full
 </pre>
 </div>
 
@@ -233,7 +238,7 @@ loop_
 
 <!-- prettier-ignore-end -->
 
-### 4. <span class="orange">analysis.cif</span>
+### 4. analysis / <span class="orange">analysis.cif</span>
 
 This file contains settings used for data analysis, including the choice
 of **calculation** and **fitting** engines, as well as user defined
@@ -243,8 +248,8 @@ of **calculation** and **fitting** engines, as well as user defined
 
 <div class="cif">
 <pre>
-<span class="blue"><b>_fit</b>.minimizer_type</span>          "lmfit (leastsq)"
-<span class="blue"><b>_fit</b>.mode</span>                    single
+<span class="blue"><b>_fitting</b>.mode_type</span>              single
+<span class="blue"><b>_fitting</b>.minimizer_type</span>         lmfit
 
 loop_
 <span class="green"><b>_alias</b>.label</span>
@@ -263,6 +268,9 @@ loop_
 
 <!-- prettier-ignore-end -->
 
+When a Bayesian fit stores persisted posterior or predictive arrays, the
+same `analysis/` directory also contains `results.h5`.
+
 <br>
 
 ---
diff --git a/docs/mkdocs.yml b/docs/mkdocs.yml
index c54ff9353..02f862023 100644
--- a/docs/mkdocs.yml
+++ b/docs/mkdocs.yml
@@ -193,10 +193,13 @@ nav:
   - Tutorials:
       - Tutorials: tutorials/index.md
       - Getting Started:
-          - LBCO quick load: tutorials/ed-18.ipynb
           - LBCO quick code: tutorials/ed-2.ipynb
           - LBCO basic load: tutorials/ed-1.ipynb
           - LBCO complete: tutorials/ed-3.ipynb
+      - Load Project:
+          - LBCO Single: tutorials/ed-18.ipynb
+          - Co2SiO4 Sequential: tutorials/ed-23.ipynb
+          - LBCO Bayesian: tutorials/ed-24.ipynb
       - Powder Diffraction:
           - Co2SiO4 pd-neut-cwl: tutorials/ed-5.ipynb
           - HS pd-neut-cwl: tutorials/ed-6.ipynb
@@ -213,7 +216,6 @@ nav:
           - PbSO4 NPD+XRD: tutorials/ed-4.ipynb
           - Si Bragg+PDF: tutorials/ed-16.ipynb
           - Co2SiO4 T-scan: tutorials/ed-17.ipynb
-          - Co2SiO4 T-scan resumed: tutorials/ed-23.ipynb
       - Simulated Data:
           - LBCO+Si McStas: tutorials/ed-9.ipynb
           - BEER McStas: tutorials/ed-20.ipynb
diff --git a/pixi.lock b/pixi.lock
index 927dd5686..974cbe77e 100644
--- a/pixi.lock
+++ b/pixi.lock
@@ -8150,6 +8150,7 @@ packages:
   - diffpy-pdffit2
   - diffpy-utils
   - gemmi
+  - h5py
   - lmfit
   - numpy
   - pandas
diff --git a/pixi.toml b/pixi.toml
index 958851cf2..70a8a310e 100644
--- a/pixi.toml
+++ b/pixi.toml
@@ -102,8 +102,8 @@ user = { features = ['py-max', 'user'] }
 unit-tests = 'python -m pytest tests/unit/ --color=yes -v'
 functional-tests = 'python -m pytest tests/functional/ --color=yes -v'
 integration-tests = 'python -m pytest tests/integration/ --color=yes -n auto -v'
-script-tests = 'python -m pytest tools/test_scripts.py --color=yes -n auto -v'
-notebook-tests = 'python -m pytest --nbmake docs/docs/tutorials/ --nbmake-timeout=1200 --color=yes -n auto -v'
+script-tests = { cmd = 'python -m pytest tools/test_scripts.py --color=yes -n auto -v', env = { EASYDIFFRACTION_ARTIFACT_ROOT = 'tmp/tutorials' } }
+notebook-tests = { cmd = 'python -m pytest --nbmake docs/docs/tutorials/ --nbmake-timeout=1200 --color=yes -n auto -v', env = { EASYDIFFRACTION_ARTIFACT_ROOT = 'tmp/tutorials' } }
 
 test = { depends-on = ['unit-tests', 'functional-tests'] }
 test-all = { depends-on = [
@@ -184,10 +184,16 @@ cov = { depends-on = [
 # 📓 Notebook Management
 ########################
 
+python = { cmd = 'python', env = { EASYDIFFRACTION_ARTIFACT_ROOT = 'tmp/tutorials' } }
+tutorial = { cmd = 'python', env = { EASYDIFFRACTION_ARTIFACT_ROOT = 'tmp/tutorials' } }
+tutorial-benchmarks = { cmd = 'python tools/benchmark_tutorials.py', env = { EASYDIFFRACTION_ARTIFACT_ROOT = 'tmp/tutorials' } }
+jupyter = { cmd = 'jupyter', env = { EASYDIFFRACTION_ARTIFACT_ROOT = 'tmp/tutorials' } }
+
 notebook-convert = 'jupytext docs/docs/tutorials/*.py --from py:percent --to ipynb'
 notebook-strip = 'nbstripout docs/docs/tutorials/*.ipynb'
 notebook-tweak = 'python tools/tweak_notebooks.py docs/docs/tutorials/'
-notebook-exec = 'python -m pytest --nbmake docs/docs/tutorials/ --nbmake-timeout=1200 --overwrite --color=yes -n auto -v'
+notebook-exec = { cmd = 'python -m pytest --nbmake docs/docs/tutorials/ --overwrite --color=yes -n auto -v', env = { EASYDIFFRACTION_ARTIFACT_ROOT = 'tmp/tutorials' } }
+notebook-exec-ci = { cmd = 'python -m pytest --nbmake docs/docs/tutorials/ --nbmake-timeout=1200 --overwrite --color=yes -n auto -v', env = { EASYDIFFRACTION_ARTIFACT_ROOT = '.' } }
 
 notebook-prepare = { depends-on = [
   'notebook-convert',
diff --git a/pyproject.toml b/pyproject.toml
index 3a4b78a01..507a069d1 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -41,6 +41,7 @@ dependencies = [
   'diffpy.pdffit2', # Calculations of Pair Distribution Function (PDF)
   'diffpy.utils',   # Utilities for PDF calculations
   'uncertainties',  # Propagation of uncertainties
+  'h5py',           # HDF5 file handling
   'typeguard',      # Runtime type checking
   'darkdetect',     # Detecting dark mode (system-level)
   'pandas',         # Displaying tables in Jupyter notebooks
diff --git a/src/easydiffraction/__init__.py b/src/easydiffraction/__init__.py
index 11ea117c8..0e69a9c7e 100644
--- a/src/easydiffraction/__init__.py
+++ b/src/easydiffraction/__init__.py
@@ -14,5 +14,6 @@
 from easydiffraction.utils.utils import download_all_tutorials
 from easydiffraction.utils.utils import download_data
 from easydiffraction.utils.utils import download_tutorial
+from easydiffraction.utils.utils import list_data
 from easydiffraction.utils.utils import list_tutorials
 from easydiffraction.utils.utils import show_version
diff --git a/src/easydiffraction/__main__.py b/src/easydiffraction/__main__.py
index ca1d3f8c4..a66d0d8f8 100644
--- a/src/easydiffraction/__main__.py
+++ b/src/easydiffraction/__main__.py
@@ -1,6 +1,8 @@
 # SPDX-FileCopyrightText: 2025 EasyScience contributors <https://github.com/easyscience>
 # SPDX-License-Identifier: BSD-3-Clause
 
+from __future__ import annotations
+
 import sys
 
 # Ensure UTF-8 output on all platforms (e.g. Windows with cp1252)
@@ -14,6 +16,90 @@
 
 app = typer.Typer(add_completion=False)
 
+_MIN_PROJECT_FIRST_ARG_COUNT = 2
+_PROJECT_COMMAND_NAMES = frozenset({'fit', 'display', 'undo'})
+_GLOBAL_COMMAND_NAMES = frozenset({
+    'list-data',
+    'download-data',
+    'list-tutorials',
+    'download-tutorial',
+    'download-all-tutorials',
+    *_PROJECT_COMMAND_NAMES,
+})
+
+
+def _normalized_cli_args(args: list[str]) -> list[str]:
+    """Return CLI args rewritten to support project-first commands."""
+    if len(args) < _MIN_PROJECT_FIRST_ARG_COUNT:
+        return args
+
+    first_arg = args[0]
+    if first_arg.startswith('-') or first_arg in _GLOBAL_COMMAND_NAMES:
+        return args
+
+    if args[1] not in _PROJECT_COMMAND_NAMES:
+        return args
+
+    return [args[1], first_arg, *args[2:]]
+
+
+def _load_project(project_dir: str) -> object:
+    """Load one saved project directory."""
+    return ed.Project.load(project_dir)
+
+
+def _display_project_patterns(project: object) -> None:
+    """Render default pattern views for all experiments."""
+    for experiment in project.experiments:
+        project.display.pattern(expt_name=experiment.name)
+
+
+def _project_fit_mode(project: object) -> str | None:
+    """Return the resolved fitting mode type for one project."""
+    return getattr(project.analysis, 'fitting_mode_type', None)
+
+
+def _project_result_kind(project: object) -> str | None:
+    """Return the resolved fit result kind for one project."""
+    result_kind = getattr(getattr(project.analysis, 'fit_result', None), 'result_kind', None)
+    return getattr(result_kind, 'value', None)
+
+
+def _display_fit_outputs(project: object) -> None:
+    """Render the standard post-fit CLI outputs."""
+    if _project_fit_mode(project) != 'sequential':
+        project.display.fit.results()
+        project.display.fit.correlations()
+    _display_project_patterns(project)
+
+
+def _display_project_outputs(project: object) -> None:
+    """Render the typical displays for the loaded project state."""
+    if _project_fit_mode(project) == 'sequential':
+        project.display.fit.series()
+        _display_project_patterns(project)
+        return
+
+    project.display.fit.results()
+    project.display.fit.correlations()
+
+    if _project_result_kind(project) == 'bayesian':
+        if project.rendering.plotter.engine == 'plotly':
+            project.display.posterior.pairs()
+        project.display.posterior.distribution()
+        for experiment in project.experiments:
+            project.display.posterior.predictive(expt_name=experiment.name)
+
+    _display_project_patterns(project)
+
+
+def run_cli(args: list[str] | None = None) -> None:
+    """
+    Run the EasyDiffraction CLI with project-first argument support.
+    """
+    cli_args = list(sys.argv[1:] if args is None else args)
+    app(args=_normalized_cli_args(cli_args))
+
 
 @app.callback(invoke_without_command=True)
 def main(
@@ -37,12 +123,38 @@ def main(
     # Otherwise, let the chosen subcommand execute.
 
 
+@app.command('list-data')
+def list_data() -> None:
+    """List available example data and project archives."""
+    ed.list_data()
+
+
 @app.command('list-tutorials')
 def list_tutorials() -> None:
     """List available tutorial notebooks."""
     ed.list_tutorials()
 
 
+@app.command('download-data')
+def download_data(
+    id: int = typer.Argument(..., help='Data ID to download.'),
+    destination: str = typer.Option(
+        'data',
+        '--destination',
+        '-d',
+        help='Directory to save the data or extracted project into.',
+    ),
+    overwrite: bool = typer.Option(  # noqa: FBT001
+        False,  # noqa: FBT003
+        '--overwrite',
+        '-o',
+        help='Overwrite an existing file or extracted project if present.',
+    ),
+) -> None:
+    """Download one example data record by ID."""
+    ed.download_data(id=id, destination=destination, overwrite=overwrite)
+
+
 @app.command('download-tutorial')
 def download_tutorial(
     id: int = typer.Argument(..., help='Tutorial ID to download.'),
@@ -82,6 +194,18 @@ def download_all_tutorials(
     ed.download_all_tutorials(destination=destination, overwrite=overwrite)
 
 
+@app.command('display')
+def display(
+    project_dir: str = typer.Argument(
+        ...,
+        help='Path to the project directory (must contain project.cif).',
+    ),
+) -> None:
+    """Display the typical outputs for a saved project state."""
+    project = _load_project(project_dir)
+    _display_project_outputs(project)
+
+
 @app.command('fit')
 def fit(
     project_dir: str = typer.Argument(
@@ -94,18 +218,28 @@ def fit(
         help='Run fitting without saving results back to the project directory.',
     ),
 ) -> None:
-    """Fit a saved project: easydiffraction fit PROJECT_DIR [--dry]."""
-    project = ed.Project.load(project_dir)
+    """Fit a saved project: easydiffraction PROJECT_DIR fit [--dry]."""
+    project = _load_project(project_dir)
     if dry:
         project.info._path = None
     project.analysis.fit()
-    if getattr(project.analysis, 'fitting_mode_type', None) != 'sequential':
-        project.display.fit.results()
-        project.display.fit.correlations()
-    for expt in project.experiments:
-        project.display.pattern(expt_name=expt.name)
-    # project.summary.show_report()
+    _display_fit_outputs(project)
+
+
+@app.command('undo')
+def undo(
+    project_dir: str = typer.Argument(
+        ...,
+        help='Path to the project directory (must contain project.cif).',
+    ),
+) -> None:
+    """
+    Undo the last fit when fit-history support exists (not implemented).
+    """
+    _load_project(project_dir)
+    typer.echo('Undo is not yet implemented.')
+    raise typer.Exit(code=1)
 
 
 if __name__ == '__main__':
-    app()
+    run_cli()
diff --git a/src/easydiffraction/analysis/__init__.py b/src/easydiffraction/analysis/__init__.py
index 0fe4386c4..f9d5e94de 100644
--- a/src/easydiffraction/analysis/__init__.py
+++ b/src/easydiffraction/analysis/__init__.py
@@ -1,6 +1,56 @@
 # SPDX-FileCopyrightText: 2025 EasyScience contributors <https://github.com/easyscience>
 # SPDX-License-Identifier: BSD-3-Clause
 
+from easydiffraction.analysis.categories.bayesian_convergence import BayesianConvergence
+from easydiffraction.analysis.categories.bayesian_convergence import BayesianConvergenceFactory
+from easydiffraction.analysis.categories.bayesian_distribution_caches import (
+    BayesianDistributionCacheItem,
+)
+from easydiffraction.analysis.categories.bayesian_distribution_caches import (
+    BayesianDistributionCaches,
+)
+from easydiffraction.analysis.categories.bayesian_distribution_caches import (
+    BayesianDistributionCachesFactory,
+)
+from easydiffraction.analysis.categories.bayesian_pair_caches import BayesianPairCacheItem
+from easydiffraction.analysis.categories.bayesian_pair_caches import BayesianPairCaches
+from easydiffraction.analysis.categories.bayesian_pair_caches import BayesianPairCachesFactory
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors import (
+    BayesianParameterPosteriorItem,
+)
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors import (
+    BayesianParameterPosteriors,
+)
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors import (
+    BayesianParameterPosteriorsFactory,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets import (
+    BayesianPredictiveDatasetItem,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets import (
+    BayesianPredictiveDatasets,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets import (
+    BayesianPredictiveDatasetsFactory,
+)
+from easydiffraction.analysis.categories.bayesian_result import BayesianResult
+from easydiffraction.analysis.categories.bayesian_result import BayesianResultFactory
+from easydiffraction.analysis.categories.bayesian_sampler import BayesianSampler
+from easydiffraction.analysis.categories.bayesian_sampler import BayesianSamplerFactory
+from easydiffraction.analysis.categories.deterministic_result import DeterministicResult
+from easydiffraction.analysis.categories.deterministic_result import DeterministicResultFactory
+from easydiffraction.analysis.categories.fit_parameter_correlations import (
+    FitParameterCorrelationItem,
+)
+from easydiffraction.analysis.categories.fit_parameter_correlations import FitParameterCorrelations
+from easydiffraction.analysis.categories.fit_parameter_correlations import (
+    FitParameterCorrelationsFactory,
+)
+from easydiffraction.analysis.categories.fit_parameters import FitParameterItem
+from easydiffraction.analysis.categories.fit_parameters import FitParameters
+from easydiffraction.analysis.categories.fit_parameters import FitParametersFactory
+from easydiffraction.analysis.categories.fit_result import FitResult
+from easydiffraction.analysis.categories.fit_result import FitResultFactory
 from easydiffraction.analysis.categories.fitting import Fitting
 from easydiffraction.analysis.categories.fitting import FittingFactory
 from easydiffraction.analysis.categories.joint_fit import JointFitCollection
@@ -13,4 +63,6 @@
 )
 from easydiffraction.analysis.categories.sequential_fit_extract import SequentialFitExtractFactory
 from easydiffraction.analysis.categories.sequential_fit_extract import SequentialFitExtractItem
+from easydiffraction.analysis.enums import FitCorrelationSourceEnum
 from easydiffraction.analysis.enums import FitModeEnum
+from easydiffraction.analysis.enums import FitResultKindEnum
diff --git a/src/easydiffraction/analysis/analysis.py b/src/easydiffraction/analysis/analysis.py
index 08303f65b..eccbf28d8 100644
--- a/src/easydiffraction/analysis/analysis.py
+++ b/src/easydiffraction/analysis/analysis.py
@@ -4,13 +4,35 @@
 from __future__ import annotations
 
 from contextlib import suppress
+from itertools import combinations
 from pathlib import Path
 
 import numpy as np
 import pandas as pd
 
 from easydiffraction.analysis.categories.aliases.factory import AliasesFactory
+from easydiffraction.analysis.categories.bayesian_convergence import BayesianConvergence
+from easydiffraction.analysis.categories.bayesian_distribution_caches import (
+    BayesianDistributionCaches,
+)
+from easydiffraction.analysis.categories.bayesian_pair_caches import BayesianPairCaches
+from easydiffraction.analysis.categories.bayesian_pair_caches.default import BayesianPairCachePaths
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors import (
+    BayesianParameterPosteriors,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets import (
+    BayesianPredictiveDatasets,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets.default import (
+    BayesianPredictiveDatasetPaths,
+)
+from easydiffraction.analysis.categories.bayesian_result import BayesianResult
+from easydiffraction.analysis.categories.bayesian_sampler import BayesianSampler
 from easydiffraction.analysis.categories.constraints.factory import ConstraintsFactory
+from easydiffraction.analysis.categories.deterministic_result import DeterministicResult
+from easydiffraction.analysis.categories.fit_parameter_correlations import FitParameterCorrelations
+from easydiffraction.analysis.categories.fit_parameters import FitParameters
+from easydiffraction.analysis.categories.fit_result import FitResult
 from easydiffraction.analysis.categories.fitting import Fitting
 from easydiffraction.analysis.categories.fitting import FittingFactory
 from easydiffraction.analysis.categories.joint_fit import JointFitCollection
@@ -19,14 +41,23 @@
 from easydiffraction.analysis.categories.sequential_fit_extract import (
     SequentialFitExtractCollection,
 )
+from easydiffraction.analysis.enums import FitCorrelationSourceEnum
 from easydiffraction.analysis.enums import FitModeEnum
+from easydiffraction.analysis.enums import FitResultKindEnum
+from easydiffraction.analysis.fit_helpers.bayesian import BayesianFitResults
+from easydiffraction.analysis.fit_helpers.bayesian import PosteriorParameterSummary
+from easydiffraction.analysis.fit_helpers.bayesian import PosteriorPredictiveSummary
+from easydiffraction.analysis.fit_helpers.bayesian import PosteriorSamples
+from easydiffraction.analysis.fit_helpers.reporting import FitResults
 from easydiffraction.analysis.fitting import Fitter
+from easydiffraction.analysis.minimizers.enums import MinimizerTypeEnum
 from easydiffraction.core.category_owner import CategoryOwner
 from easydiffraction.core.guard import _apply_help_filter
 from easydiffraction.core.singleton import ConstraintsHandler
 from easydiffraction.core.variable import NumericDescriptor
 from easydiffraction.core.variable import Parameter
 from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.datablocks.experiment.item.base import intensity_category_for
 from easydiffraction.display.progress import make_display_handle
 from easydiffraction.display.tables import TableRenderer
 from easydiffraction.io.cif.serialize import analysis_to_cif
@@ -40,6 +71,9 @@
 from easydiffraction.utils.utils import render_table
 
 _SUMMARY_HIDDEN_PARAMETER_CATEGORIES = frozenset({'pd_data', 'total_data', 'refln'})
+_POSTERIOR_SAMPLE_NDIM = 3
+_FLATTENED_POSTERIOR_SAMPLE_NDIM = 2
+_CREDIBLE_INTERVAL_LEVEL_COUNT = 2
 
 
 def _discover_property_rows(cls: type) -> list[list[str]]:
@@ -347,7 +381,111 @@ def as_cif(self) -> None:
         self._analysis.show_as_cif()
 
 
-class Analysis(CategoryOwner):
+class _AnalysisOwnerAccessorsMixin:
+    @property
+    def project(self) -> object:
+        """Project that owns this analysis section."""
+        return self._project
+
+    @property
+    def aliases(self) -> object:
+        """Alias mappings used by symbolic constraints and displays."""
+        return self._aliases
+
+    @property
+    def constraints(self) -> object:
+        """Symbolic constraints owned by this analysis section."""
+        return self._constraints
+
+    @property
+    def display(self) -> AnalysisDisplay:
+        """Display helper for parameter tables, CIF, and fit results."""
+        return self._display
+
+    @property
+    def fitter(self) -> Fitter:
+        """Fitting engine used by this analysis object."""
+        return self._fitter
+
+    @fitter.setter
+    def fitter(self, value: Fitter) -> None:
+        self._fitter = value
+
+    @property
+    def fit_results(self) -> object | None:
+        """Results from the most recent fit, if any."""
+        if self._fit_results is None and self._has_persisted_fit_state():
+            self._restore_fit_results_from_projection()
+        return self._fit_results
+
+    @fit_results.setter
+    def fit_results(self, value: object | None) -> None:
+        self._fit_results = value
+        self._fitter.results = value
+
+
+class _AnalysisPersistedCategoryAccessorsMixin:
+    @property
+    def fit_parameters(self) -> FitParameters:
+        """Persisted fit-parameter control snapshots."""
+        return self._fit_parameters
+
+    @property
+    def fit_result(self) -> FitResult:
+        """Persisted common fit-result status metadata."""
+        return self._fit_result
+
+    @property
+    def fit_parameter_correlations(self) -> FitParameterCorrelations:
+        """Persisted fit-parameter correlation summaries."""
+        return self._fit_parameter_correlations
+
+    @property
+    def deterministic_result(self) -> DeterministicResult:
+        """Persisted deterministic fit-result metadata."""
+        return self._deterministic_result
+
+    @property
+    def bayesian_result(self) -> BayesianResult:
+        """Persisted Bayesian fit-result metadata."""
+        return self._bayesian_result
+
+    @property
+    def bayesian_sampler(self) -> BayesianSampler:
+        """Persisted Bayesian sampler settings."""
+        return self._bayesian_sampler
+
+    @property
+    def bayesian_convergence(self) -> BayesianConvergence:
+        """Persisted Bayesian convergence diagnostics."""
+        return self._bayesian_convergence
+
+    @property
+    def bayesian_parameter_posteriors(self) -> BayesianParameterPosteriors:
+        """Persisted Bayesian parameter posterior summaries."""
+        return self._bayesian_parameter_posteriors
+
+    @property
+    def bayesian_distribution_caches(self) -> BayesianDistributionCaches:
+        """Persisted Bayesian distribution-cache manifests."""
+        return self._bayesian_distribution_caches
+
+    @property
+    def bayesian_pair_caches(self) -> BayesianPairCaches:
+        """Persisted Bayesian pair-cache manifests."""
+        return self._bayesian_pair_caches
+
+    @property
+    def bayesian_predictive_datasets(self) -> BayesianPredictiveDatasets:
+        """Persisted Bayesian predictive-dataset manifests."""
+        return self._bayesian_predictive_datasets
+
+
+class Analysis(
+    _AnalysisOwnerAccessorsMixin,
+    _AnalysisPersistedCategoryAccessorsMixin,
+    CategoryOwner,
+):
     """
     High-level orchestration of analysis tasks for a Project.
 
@@ -379,48 +517,321 @@ def __init__(self, project: object) -> None:
             SequentialFitFactory.default_tag()
         )
         self._sequential_fit_extract = SequentialFitExtractCollection()
+        self._fit_parameters = FitParameters()
+        self._fit_result = FitResult()
+        self._fit_parameter_correlations = FitParameterCorrelations()
+        self._deterministic_result = DeterministicResult()
+        self._bayesian_result = BayesianResult()
+        self._bayesian_sampler = BayesianSampler()
+        self._bayesian_convergence = BayesianConvergence()
+        self._bayesian_parameter_posteriors = BayesianParameterPosteriors()
+        self._bayesian_distribution_caches = BayesianDistributionCaches()
+        self._bayesian_pair_caches = BayesianPairCaches()
+        self._bayesian_predictive_datasets = BayesianPredictiveDatasets()
+        self._has_persisted_fit_state_data = False
+        self._persisted_fit_state_sidecar: dict[str, object] = {}
         self._fitter = Fitter(self._fitting.minimizer_type.value)
         self._fit_results = None
         self._parameter_snapshots: dict[str, dict[str, dict]] = {}
         self._display = AnalysisDisplay(self)
 
-    @property
-    def project(self) -> object:
-        """Project that owns this analysis section."""
-        return self._project
+    @staticmethod
+    def _predictive_cache_key(
+        experiment_name: str,
+        x_axis_name: str,
+        *,
+        include_draws: bool = True,
+    ) -> str:
+        """Return the runtime cache key for one predictive summary."""
+        key_suffix = 'draws' if include_draws else 'band'
+        return f'{experiment_name}:{x_axis_name}:{key_suffix}'
+
+    def _live_parameter_map(self) -> dict[str, Parameter]:
+        """Return live parameters keyed by unique name."""
+        all_parameters = self.project.structures.parameters + self.project.experiments.parameters
+        return {
+            param.unique_name: param
+            for param in all_parameters
+            if isinstance(param, Parameter) and hasattr(param, 'unique_name')
+        }
 
-    @property
-    def aliases(self) -> object:
-        """Alias mappings used by symbolic constraints and displays."""
-        return self._aliases
+    def _ordered_restored_parameter_names(self) -> list[str]:
+        """
+        Return persisted parameter names in display and array order.
+        """
+        if self.fit_result.result_kind.value == FitResultKindEnum.BAYESIAN.value:
+            posterior_rows = list(self.bayesian_parameter_posteriors)
+            if posterior_rows:
+                return [row.unique_name.value for row in posterior_rows]
+
+        return [row.param_unique_name.value for row in self.fit_parameters]
+
+    def _restore_live_parameter_state(self, param_map: dict[str, Parameter]) -> None:
+        """Restore saved fit metadata onto live parameter objects."""
+        for row in self.fit_parameters:
+            parameter = param_map.get(row.param_unique_name.value)
+            if parameter is None:
+                log.warning(
+                    'Persisted fit-state references unknown parameter '
+                    f'{row.param_unique_name.value!r}.'
+                )
+                continue
 
-    @property
-    def constraints(self) -> object:
-        """Symbolic constraints owned by this analysis section."""
-        return self._constraints
+            parameter.fit_min = row.fit_min.value
+            parameter.fit_max = row.fit_max.value
+            parameter._set_fit_bounds_uncertainty_multiplier(
+                row.fit_bounds_uncertainty_multiplier.value
+            )
+            parameter._fit_start_value = row.start_value.value
+            parameter._fit_start_uncertainty = row.start_uncertainty.value
+
+        for row in self.bayesian_parameter_posteriors:
+            parameter = param_map.get(row.unique_name.value)
+            if parameter is None or row.uncertainty.value is None:
+                continue
+            parameter.uncertainty = float(row.uncertainty.value)
+
+    def _sync_live_minimizer_from_persisted_fit_state(self) -> None:
+        """Apply saved sampler settings to the live minimizer."""
+        if not self._has_persisted_fit_state():
+            return
 
-    @property
-    def display(self) -> AnalysisDisplay:
-        """Display helper for parameter tables, CIF, and fit results."""
-        return self._display
+        if self.fit_result.result_kind.value != FitResultKindEnum.BAYESIAN.value:
+            return
 
-    @property
-    def fitter(self) -> Fitter:
-        """Fitting engine used by this analysis object."""
-        return self._fitter
+        if self.fitting.minimizer_type.value != MinimizerTypeEnum.BUMPS_DREAM.value:
+            return
 
-    @fitter.setter
-    def fitter(self, value: Fitter) -> None:
-        self._fitter = value
+        minimizer = self.fitting.minimizer
+        if minimizer is None:
+            return
 
-    @property
-    def fit_results(self) -> object | None:
-        """Results from the most recent fit, if any."""
-        return self._fit_results
+        steps = int(self.bayesian_sampler.steps.value)
+        if steps <= 0:
+            return
 
-    @fit_results.setter
-    def fit_results(self, value: object | None) -> None:
-        self._fit_results = value
+        minimizer.steps = steps
+        minimizer.burn = int(self.bayesian_sampler.burn.value)
+
+        thin = int(self.bayesian_sampler.thin.value)
+        if thin > 0:
+            minimizer.thin = thin
+
+        pop = int(self.bayesian_sampler.pop.value)
+        if pop > 0:
+            minimizer.pop = pop
+
+        minimizer.parallel = int(self.bayesian_sampler.parallel.value)
+
+        init_value = str(self.bayesian_sampler.init.value)
+        if init_value:
+            minimizer.init = init_value
+
+    def _restored_fit_parameters(self, param_map: dict[str, Parameter]) -> list[Parameter]:
+        """Return live parameters in the persisted fit-result order."""
+        restored_parameters: list[Parameter] = []
+        for unique_name in self._ordered_restored_parameter_names():
+            parameter = param_map.get(unique_name)
+            if parameter is not None:
+                restored_parameters.append(parameter)
+        return restored_parameters
+
+    def _restored_posterior_samples(self) -> PosteriorSamples | None:
+        """Return restored posterior samples from the HDF5 sidecar."""
+        if not self.bayesian_result.has_posterior_samples.value:
+            return None
+
+        posterior_data = self._persisted_fit_state_sidecar.get('posterior', {})
+        parameter_samples = posterior_data.get('parameter_samples')
+        if parameter_samples is None:
+            return None
+
+        posterior_rows = list(self.bayesian_parameter_posteriors)
+        parameter_names = [row.unique_name.value for row in posterior_rows]
+        if not parameter_names:
+            parameter_names = [row.param_unique_name.value for row in self.fit_parameters]
+
+        parameter_sample_array = np.asarray(parameter_samples, dtype=float)
+        if parameter_sample_array.ndim != _POSTERIOR_SAMPLE_NDIM:
+            log.warning('Persisted posterior samples have an invalid shape for restore.')
+            return None
+        if parameter_sample_array.shape[2] != len(parameter_names):
+            log.warning(
+                'Persisted posterior samples do not match restored posterior parameter names.'
+            )
+            return None
+
+        log_posterior = posterior_data.get('log_posterior')
+        draw_index = posterior_data.get('draw_index')
+        return PosteriorSamples(
+            parameter_names=parameter_names,
+            parameter_samples=parameter_sample_array,
+            log_posterior=(
+                None if log_posterior is None else np.asarray(log_posterior, dtype=float)
+            ),
+            draw_index=None if draw_index is None else np.asarray(draw_index),
+        )
+
+    def _restored_posterior_summaries(self) -> list[PosteriorParameterSummary]:
+        """Return posterior summary rows as runtime summary objects."""
+        return [
+            PosteriorParameterSummary(
+                unique_name=row.unique_name.value,
+                display_name=row.display_name.value,
+                best_sample_value=float(row.best_sample_value.value),
+                median=float(row.median.value),
+                standard_deviation=float(row.uncertainty.value),
+                interval_68=(
+                    float(row.interval_68_lower.value),
+                    float(row.interval_68_upper.value),
+                ),
+                interval_95=(
+                    float(row.interval_95_lower.value),
+                    float(row.interval_95_upper.value),
+                ),
+                ess_bulk=row.ess_bulk.value,
+                r_hat=row.r_hat.value,
+            )
+            for row in self.bayesian_parameter_posteriors
+        ]
+
+    def _restored_predictive_summaries(self) -> dict[str, PosteriorPredictiveSummary]:
+        """Return restored predictive summaries for runtime reuse."""
+        restored_predictive: dict[str, PosteriorPredictiveSummary] = {}
+        predictive_data = self._persisted_fit_state_sidecar.get('predictive_datasets', {})
+        for row in self.bayesian_predictive_datasets:
+            experiment_name = str(row.experiment_name.value)
+            x_axis_name = str(row.x_axis_name.value)
+            dataset = predictive_data.get(experiment_name)
+            if dataset is None:
+                continue
+
+            summary = PosteriorPredictiveSummary(
+                experiment_name=experiment_name,
+                x_axis_name=x_axis_name,
+                x=np.asarray(dataset['x'], dtype=float),
+                best_sample_prediction=np.asarray(
+                    dataset['best_sample_prediction'],
+                    dtype=float,
+                ),
+                lower_95=(
+                    None
+                    if dataset.get('lower_95') is None
+                    else np.asarray(dataset['lower_95'], dtype=float)
+                ),
+                upper_95=(
+                    None
+                    if dataset.get('upper_95') is None
+                    else np.asarray(dataset['upper_95'], dtype=float)
+                ),
+                lower_68=(
+                    None
+                    if dataset.get('lower_68') is None
+                    else np.asarray(dataset['lower_68'], dtype=float)
+                ),
+                upper_68=(
+                    None
+                    if dataset.get('upper_68') is None
+                    else np.asarray(dataset['upper_68'], dtype=float)
+                ),
+                draws=(
+                    None
+                    if dataset.get('draws') is None
+                    else np.asarray(dataset['draws'], dtype=float)
+                ),
+            )
+            restored_predictive[experiment_name] = summary
+            restored_predictive[
+                self._predictive_cache_key(
+                    experiment_name,
+                    x_axis_name,
+                    include_draws=False,
+                )
+            ] = summary
+            if summary.draws is not None:
+                restored_predictive[
+                    self._predictive_cache_key(
+                        experiment_name,
+                        x_axis_name,
+                        include_draws=True,
+                    )
+                ] = summary
+        return restored_predictive
+
+    def _restore_fit_results_from_projection(self) -> object | None:
+        """Rebuild a runtime fit-result object from saved state."""
+        if not self._has_persisted_fit_state():
+            return None
+
+        param_map = self._live_parameter_map()
+        self._restore_live_parameter_state(param_map)
+        restored_parameters = self._restored_fit_parameters(param_map)
+        fitting_time = self.fit_result.fitting_time.value
+        reduced_chi_square = self.fit_result.reduced_chi_square.value
+
+        if self.fit_result.result_kind.value == FitResultKindEnum.BAYESIAN.value:
+            restored_results = BayesianFitResults(
+                success=bool(self.fit_result.success.value),
+                parameters=restored_parameters,
+                reduced_chi_square=reduced_chi_square,
+                starting_parameters=list(restored_parameters),
+                fitting_time=fitting_time,
+                sampler_name=self.bayesian_result.sampler_name.value,
+                point_estimate_name=self.bayesian_result.point_estimate_name.value,
+                posterior_samples=self._restored_posterior_samples(),
+                posterior_parameter_summaries=self._restored_posterior_summaries(),
+                posterior_predictive=self._restored_predictive_summaries(),
+                credible_interval_levels=(
+                    float(self.bayesian_result.credible_interval_inner.value),
+                    float(self.bayesian_result.credible_interval_outer.value),
+                ),
+                sampler_settings={
+                    'steps': int(self.bayesian_sampler.steps.value),
+                    'burn': int(self.bayesian_sampler.burn.value),
+                    'thin': int(self.bayesian_sampler.thin.value),
+                    'pop': int(self.bayesian_sampler.pop.value),
+                    'parallel': int(self.bayesian_sampler.parallel.value),
+                    'init': self.bayesian_sampler.init.value,
+                    'random_seed': self.bayesian_sampler.random_seed.value,
+                },
+                convergence_diagnostics={
+                    'converged': bool(self.bayesian_convergence.converged.value),
+                    'max_r_hat': self.bayesian_convergence.max_r_hat.value,
+                    'min_ess_bulk': self.bayesian_convergence.min_ess_bulk.value,
+                    'n_draws': int(self.bayesian_convergence.n_draws.value),
+                    'n_chains': int(self.bayesian_convergence.n_chains.value),
+                    'n_parameters': int(self.bayesian_convergence.n_parameters.value),
+                },
+                sampler_completed=bool(self.bayesian_result.sampler_completed.value),
+                best_log_posterior=self.bayesian_result.best_log_posterior.value,
+            )
+            restored_results.message = self.fit_result.message.value
+            restored_results.iterations = int(self.fit_result.iterations.value)
+            self.fit_results = restored_results
+            return restored_results
+
+        restored_results = FitResults(
+            success=bool(self.fit_result.success.value),
+            parameters=restored_parameters,
+            reduced_chi_square=reduced_chi_square,
+            starting_parameters=list(restored_parameters),
+            fitting_time=fitting_time,
+            optimizer_name=self.deterministic_result.optimizer_name.value,
+            method_name=self.deterministic_result.method_name.value,
+            objective_name=self.deterministic_result.objective_name.value,
+            objective_value=self.deterministic_result.objective_value.value,
+            n_data_points=int(self.deterministic_result.n_data_points.value),
+            n_parameters=int(self.deterministic_result.n_parameters.value),
+            n_free_parameters=int(self.deterministic_result.n_free_parameters.value),
+            degrees_of_freedom=int(self.deterministic_result.degrees_of_freedom.value),
+            covariance_available=bool(self.deterministic_result.covariance_available.value),
+            correlation_available=bool(self.deterministic_result.correlation_available.value),
+        )
+        restored_results.message = self.fit_result.message.value
+        restored_results.iterations = int(self.fit_result.iterations.value)
+        restored_results.chi_square = self.deterministic_result.objective_value.value
+        self.fit_results = restored_results
+        return restored_results
 
     def help(self) -> None:
         """Print a summary of analysis properties and methods."""
@@ -502,6 +913,9 @@ def _serializable_categories(self) -> list:
                 self.sequential_fit_extract,
             ])
 
+        if self._has_persisted_fit_state():
+            categories.extend(self._fit_state_categories())
+
         return categories
 
     # ------------------------------------------------------------------
@@ -684,6 +1098,659 @@ def sequential_fit_extract(self) -> SequentialFitExtractCollection:
         """Persisted extract rules for sequential fitting."""
         return self._sequential_fit_extract
 
+    def _has_persisted_fit_state(self) -> bool:
+        """
+        Return whether a persisted fit-state projection is present.
+        """
+        return self._has_persisted_fit_state_data
+
+    def _set_has_persisted_fit_state(self, *, value: bool) -> None:
+        """Set the persisted fit-state presence flag."""
+        self._has_persisted_fit_state_data = value
+
+    def _fit_state_categories(self) -> list[object]:
+        """Return fit-state categories for the current result kind."""
+        categories: list[object] = [
+            self.fit_parameters,
+            self.fit_result,
+            self.fit_parameter_correlations,
+        ]
+
+        try:
+            result_kind = FitResultKindEnum(self.fit_result.result_kind.value)
+        except ValueError:
+            log.warning(
+                'Unsupported fit_result.result_kind while serializing analysis CIF: '
+                f'{self.fit_result.result_kind.value!r}. '
+                'Saving only common fit-state categories.',
+            )
+            return categories
+
+        if result_kind is FitResultKindEnum.DETERMINISTIC:
+            categories.append(self.deterministic_result)
+            return categories
+
+        categories.extend([
+            self.bayesian_result,
+            self.bayesian_sampler,
+            self.bayesian_convergence,
+            self.bayesian_parameter_posteriors,
+            self.bayesian_distribution_caches,
+            self.bayesian_pair_caches,
+            self.bayesian_predictive_datasets,
+        ])
+        return categories
+
+    def _clear_persisted_fit_state(self) -> None:
+        """Reset all persisted fit-state categories before a new fit."""
+        self._fit_parameters = FitParameters()
+        self._fit_result = FitResult()
+        self._fit_parameter_correlations = FitParameterCorrelations()
+        self._deterministic_result = DeterministicResult()
+        self._bayesian_result = BayesianResult()
+        self._bayesian_sampler = BayesianSampler()
+        self._bayesian_convergence = BayesianConvergence()
+        self._bayesian_parameter_posteriors = BayesianParameterPosteriors()
+        self._bayesian_distribution_caches = BayesianDistributionCaches()
+        self._bayesian_pair_caches = BayesianPairCaches()
+        self._bayesian_predictive_datasets = BayesianPredictiveDatasets()
+        self._set_has_persisted_fit_state(value=False)
+        self._persisted_fit_state_sidecar = {}
+
+    def _capture_fit_parameter_state(self, parameters: list[Parameter]) -> None:
+        """Capture pre-fit parameter state."""
+        self._clear_persisted_fit_state()
+
+        for param in parameters:
+            self.fit_parameters.create(
+                param_unique_name=param.unique_name,
+                fit_min=param.fit_min,
+                fit_max=param.fit_max,
+                fit_bounds_uncertainty_multiplier=param.fit_bounds_uncertainty_multiplier,
+                start_value=param.value,
+                start_uncertainty=param.uncertainty,
+            )
+
+        self._set_has_persisted_fit_state(value=True)
+
+    def _selected_parameters_for_fit(self, experiments: list[object]) -> list[Parameter]:
+        """
+        Return unique live parameters involved in the current fit slice.
+        """
+        selected_parameters: list[Parameter] = []
+        seen_unique_names: set[str] = set()
+
+        for param in self.project.structures.parameters:
+            if not isinstance(param, Parameter):
+                continue
+            if param.unique_name in seen_unique_names:
+                continue
+            selected_parameters.append(param)
+            seen_unique_names.add(param.unique_name)
+
+        for experiment in experiments:
+            for param in experiment.parameters:
+                if not isinstance(param, Parameter):
+                    continue
+                if param.unique_name in seen_unique_names:
+                    continue
+                selected_parameters.append(param)
+                seen_unique_names.add(param.unique_name)
+
+        return selected_parameters
+
+    @staticmethod
+    def _fit_data_point_count(experiments: list[object]) -> int:
+        """Return observed data-point count for one fit slice."""
+        total = 0
+        for experiment in experiments:
+            intensity_category = intensity_category_for(experiment)
+            total += int(np.asarray(intensity_category.intensity_meas).size)
+        return total
+
+    @staticmethod
+    def _resolve_covariance_matrix(results: FitResults) -> np.ndarray | None:
+        """
+        Return a covariance matrix when the raw fit result exposes one.
+        """
+        raw_result = results.engine_result
+        for attribute_name in ('covar', 'covariance_matrix'):
+            covariance = getattr(raw_result, attribute_name, None)
+            if covariance is None:
+                continue
+
+            covariance_array = np.asarray(covariance, dtype=float)
+            if covariance_array.ndim != _FLATTENED_POSTERIOR_SAMPLE_NDIM:
+                continue
+            if covariance_array.shape[0] != covariance_array.shape[1]:
+                continue
+            return covariance_array
+
+        return None
+
+    @staticmethod
+    def _correlation_matrix_from_covariance(covariance: np.ndarray) -> np.ndarray | None:
+        """
+        Return a correlation matrix derived from a covariance matrix.
+        """
+        diagonal = np.diag(covariance)
+        if np.any(diagonal <= 0):
+            return None
+
+        scales = np.sqrt(diagonal)
+        denominator = np.outer(scales, scales)
+        with np.errstate(invalid='ignore', divide='ignore'):
+            correlation = covariance / denominator
+
+        if not np.all(np.isfinite(correlation)):
+            return None
+        return correlation
+
+    @staticmethod
+    def _resolve_objective_value(results: FitResults) -> float | None:
+        """Return the objective value stored for a fit result."""
+        if results.chi_square is None:
+            return None
+        return float(results.chi_square)
+
+    def _store_common_fit_result_projection(
+        self,
+        results: FitResults,
+        *,
+        result_kind: FitResultKindEnum,
+    ) -> None:
+        """
+        Store fields shared by deterministic and Bayesian fit results.
+        """
+        self.fit_result._set_result_kind(result_kind.value)
+        self.fit_result._set_success(value=results.success)
+        self.fit_result._set_message(results.message)
+        self.fit_result._set_iterations(results.iterations)
+        self.fit_result._set_fitting_time(results.fitting_time)
+        self.fit_result._set_reduced_chi_square(results.reduced_chi_square)
+        self._set_has_persisted_fit_state(value=True)
+
+    def _store_correlation_projection(
+        self,
+        *,
+        unique_names: list[str],
+        correlation_matrix: np.ndarray,
+        source_kind: FitCorrelationSourceEnum,
+    ) -> None:
+        """Store upper-triangle correlations from one matrix."""
+        if len(unique_names) <= 1:
+            return
+        if correlation_matrix.shape != (len(unique_names), len(unique_names)):
+            return
+
+        for row_index, unique_name_i in enumerate(unique_names[:-1]):
+            for column_index in range(row_index + 1, len(unique_names)):
+                correlation = correlation_matrix[row_index, column_index]
+                if not np.isfinite(correlation):
+                    continue
+                self.fit_parameter_correlations.create(
+                    source_kind=source_kind.value,
+                    param_unique_name_i=unique_name_i,
+                    param_unique_name_j=unique_names[column_index],
+                    correlation=float(np.clip(correlation, -1.0, 1.0)),
+                )
+
+    def _store_deterministic_result_projection(
+        self,
+        results: FitResults,
+        *,
+        experiments: list[object],
+        fitted_parameters: list[Parameter],
+    ) -> None:
+        """Store deterministic fit results in persisted categories."""
+        selected_parameters = self._selected_parameters_for_fit(experiments)
+        n_parameters = len(selected_parameters)
+        n_free_parameters = len(fitted_parameters)
+        n_data_points = self._fit_data_point_count(experiments)
+        degrees_of_freedom = max(n_data_points - n_free_parameters, 0)
+        covariance = self._resolve_covariance_matrix(results)
+        correlation_matrix = (
+            self._correlation_matrix_from_covariance(covariance)
+            if covariance is not None
+            else None
+        )
+
+        self.deterministic_result._set_optimizer_name(
+            str(self.fitter.minimizer.name or self.fitter.selection)
+        )
+        self.deterministic_result._set_method_name(str(self.fitter.minimizer.method or ''))
+        self.deterministic_result._set_objective_name('chi_square')
+        self.deterministic_result._set_objective_value(self._resolve_objective_value(results))
+        self.deterministic_result._set_n_data_points(n_data_points)
+        self.deterministic_result._set_n_parameters(n_parameters)
+        self.deterministic_result._set_n_free_parameters(n_free_parameters)
+        self.deterministic_result._set_degrees_of_freedom(degrees_of_freedom)
+        self.deterministic_result._set_covariance_available(value=covariance is not None)
+        self.deterministic_result._set_correlation_available(value=correlation_matrix is not None)
+
+        if correlation_matrix is not None:
+            self._store_correlation_projection(
+                unique_names=[param.unique_name for param in fitted_parameters],
+                correlation_matrix=correlation_matrix,
+                source_kind=FitCorrelationSourceEnum.DETERMINISTIC,
+            )
+
+    def _store_bayesian_distribution_cache_projection(
+        self,
+        *,
+        plotter: object,
+        results: BayesianFitResults,
+        flattened_samples: np.ndarray,
+        parameter_names: list[str],
+    ) -> dict[str, dict[str, np.ndarray]]:
+        """
+        Store cached posterior density curves into persisted manifests.
+        """
+        payload: dict[str, dict[str, np.ndarray]] = {}
+        for parameter_index, parameter_name in enumerate(parameter_names):
+            lower_bound, upper_bound = plotter._posterior_parameter_bounds(
+                fit_results=results,
+                parameter_name=parameter_name,
+            )
+            density_curve = plotter._posterior_density_curve(
+                flattened_samples[:, parameter_index],
+                lower_bound=lower_bound,
+                upper_bound=upper_bound,
+            )
+            if density_curve is None:
+                continue
+
+            x_values, density_values = density_curve
+            x_array = np.asarray(x_values, dtype=float)
+            density_array = np.asarray(density_values, dtype=float)
+            cache_index = len(payload)
+            self.bayesian_distribution_caches.create(
+                param_unique_name=parameter_name,
+                x_path=f'/posterior/distribution/{cache_index}/x',
+                density_path=f'/posterior/distribution/{cache_index}/density',
+                n_grid=float(x_array.size),
+                n_draws_cached=float(np.isfinite(flattened_samples[:, parameter_index]).sum()),
+            )
+            payload[parameter_name] = {
+                'x': x_array,
+                'density': density_array,
+            }
+        return payload
+
+    @staticmethod
+    def _posterior_pair_contour_levels(density: np.ndarray) -> np.ndarray:
+        """Return default contour levels for one cached pair."""
+        density_max = float(np.max(density))
+        if not np.isfinite(density_max) or density_max <= 0:
+            return np.asarray([], dtype=float)
+        return density_max * np.asarray([0.20, 0.35, 0.50, 0.65, 0.80, 0.95], dtype=float)
+
+    @staticmethod
+    def _ordered_pair_metadata(
+        parameter_names: list[str],
+        first_index: int,
+        second_index: int,
+    ) -> tuple[int, int, str, str]:
+        """Return ordered pair indices and parameter names."""
+        x_index = first_index
+        y_index = second_index
+        x_name = parameter_names[x_index]
+        y_name = parameter_names[y_index]
+        if x_name > y_name:
+            x_index, y_index = y_index, x_index
+            x_name, y_name = y_name, x_name
+        return x_index, y_index, x_name, y_name
+
+    def _store_one_bayesian_pair_cache_projection(
+        self,
+        *,
+        plotter: object,
+        results: BayesianFitResults,
+        density_samples: np.ndarray,
+        pair_metadata: tuple[int, int, str, str],
+        contour_grid_size: int,
+        pair_id: str,
+    ) -> tuple[str, dict[str, np.ndarray]] | None:
+        """Store one cached pair surface and return its payload."""
+        x_index, y_index, x_name, y_name = pair_metadata
+
+        x_values = density_samples[:, x_index]
+        y_values = density_samples[:, y_index]
+        x_bounds, y_bounds = plotter._posterior_pair_bounds(
+            fit_results=results,
+            x_parameter_name=x_name,
+            y_parameter_name=y_name,
+            x_values=x_values,
+            y_values=y_values,
+        )
+        density_surface = plotter._posterior_pair_density_surface(
+            x_values=x_values,
+            y_values=y_values,
+            x_bounds=x_bounds,
+            y_bounds=y_bounds,
+            grid_size=contour_grid_size,
+        )
+        if density_surface is None:
+            return None
+
+        x_grid_array = np.asarray(density_surface[0], dtype=float)
+        y_grid_array = np.asarray(density_surface[1], dtype=float)
+        density_array = np.asarray(density_surface[2], dtype=float)
+        contour_levels = self._posterior_pair_contour_levels(density_array)
+        self.bayesian_pair_caches.create(
+            id=pair_id,
+            parameter_names=(x_name, y_name),
+            paths=BayesianPairCachePaths(
+                x_path=f'/posterior/pairs/{pair_id}/x',
+                y_path=f'/posterior/pairs/{pair_id}/y',
+                density_path=f'/posterior/pairs/{pair_id}/density',
+                contour_level_path=f'/posterior/pairs/{pair_id}/contour_levels',
+            ),
+            grid_shape=(float(x_grid_array.size), float(y_grid_array.size)),
+            n_draws_cached=float(density_samples.shape[0]),
+        )
+        return pair_id, {
+            'x': x_grid_array,
+            'y': y_grid_array,
+            'density': density_array,
+            'contour_levels': contour_levels,
+        }
+
+    def _store_bayesian_pair_cache_projection(
+        self,
+        *,
+        plotter: object,
+        results: BayesianFitResults,
+        flattened_samples: np.ndarray,
+        parameter_names: list[str],
+    ) -> dict[str, dict[str, np.ndarray]]:
+        """Store cached pair-density surfaces in manifests."""
+        n_parameters = len(parameter_names)
+        if n_parameters <= 1:
+            return {}
+
+        density_samples = plotter._thin_posterior_samples(
+            flattened_samples,
+            max_points=plotter._posterior_pair_density_max_points(n_parameters),
+        )
+        contour_grid_size = plotter._posterior_pair_contour_grid_size(n_parameters)
+        payload: dict[str, dict[str, np.ndarray]] = {}
+        for first_index, second_index in combinations(range(n_parameters), 2):
+            pair_id = str(len(payload) + 1)
+            cache_projection = self._store_one_bayesian_pair_cache_projection(
+                plotter=plotter,
+                results=results,
+                density_samples=density_samples,
+                pair_metadata=self._ordered_pair_metadata(
+                    parameter_names,
+                    first_index,
+                    second_index,
+                ),
+                contour_grid_size=contour_grid_size,
+                pair_id=pair_id,
+            )
+            if cache_projection is None:
+                continue
+
+            pair_id, pair_payload = cache_projection
+            payload[pair_id] = pair_payload
+        return payload
+
+    @staticmethod
+    def _predictive_dataset_payload(
+        summary: PosteriorPredictiveSummary,
+    ) -> dict[str, np.ndarray]:
+        """Return persisted predictive arrays for one summary."""
+        payload: dict[str, np.ndarray] = {
+            'x': np.asarray(summary.x, dtype=float),
+            'best_sample_prediction': np.asarray(summary.best_sample_prediction, dtype=float),
+        }
+        if summary.lower_95 is not None:
+            payload['lower_95'] = np.asarray(summary.lower_95, dtype=float)
+        if summary.upper_95 is not None:
+            payload['upper_95'] = np.asarray(summary.upper_95, dtype=float)
+        if summary.lower_68 is not None:
+            payload['lower_68'] = np.asarray(summary.lower_68, dtype=float)
+        if summary.upper_68 is not None:
+            payload['upper_68'] = np.asarray(summary.upper_68, dtype=float)
+        if summary.draws is not None:
+            payload['draws'] = np.asarray(summary.draws, dtype=float)
+        return payload
+
+    def _store_bayesian_predictive_projection(
+        self,
+        *,
+        plotter: object,
+        results: BayesianFitResults,
+    ) -> dict[str, dict[str, np.ndarray]]:
+        """
+        Store posterior predictive summaries into persisted manifests.
+        """
+        predictive_payload: dict[str, dict[str, np.ndarray]] = {}
+        for experiment_name in self.project.experiments.names:
+            experiment = self.project.experiments[experiment_name]
+            x_axis, x_axis_name, _, _, _ = plotter._resolve_x_axis(experiment.type, None)
+            summary = plotter._build_posterior_predictive_summary(
+                fit_results=results,
+                experiment=experiment,
+                expt_name=experiment_name,
+                x_axis=x_axis,
+                include_draws=False,
+            )
+            if summary is None:
+                continue
+
+            results.posterior_predictive[summary.experiment_name] = summary
+            results.posterior_predictive[
+                self._predictive_cache_key(
+                    summary.experiment_name,
+                    str(x_axis_name),
+                    include_draws=False,
+                )
+            ] = summary
+            predictive_payload[summary.experiment_name] = self._predictive_dataset_payload(
+                summary,
+            )
+            predictive_root = f'/predictive/{summary.experiment_name}'
+            self.bayesian_predictive_datasets.create(
+                experiment_name=summary.experiment_name,
+                x_axis_name=str(x_axis_name),
+                paths=BayesianPredictiveDatasetPaths(
+                    x_path=f'{predictive_root}/x',
+                    best_sample_prediction_path=(f'{predictive_root}/best_sample_prediction'),
+                    lower_95_path=(
+                        None if summary.lower_95 is None else f'{predictive_root}/lower_95'
+                    ),
+                    upper_95_path=(
+                        None if summary.upper_95 is None else f'{predictive_root}/upper_95'
+                    ),
+                    lower_68_path=(
+                        None if summary.lower_68 is None else f'{predictive_root}/lower_68'
+                    ),
+                    upper_68_path=(
+                        None if summary.upper_68 is None else f'{predictive_root}/upper_68'
+                    ),
+                    draws_path=(None if summary.draws is None else f'{predictive_root}/draws'),
+                ),
+                n_x=float(np.asarray(summary.x).size),
+                n_draws_cached=(
+                    0.0 if summary.draws is None else float(np.asarray(summary.draws).shape[0])
+                ),
+            )
+        return predictive_payload
+
+    def _store_bayesian_plot_cache_projection(self, results: BayesianFitResults) -> None:
+        """Populate persisted Bayesian plot caches."""
+        posterior_samples = results.posterior_samples
+        if posterior_samples is None:
+            self._persisted_fit_state_sidecar['distribution_caches'] = {}
+            self._persisted_fit_state_sidecar['pair_caches'] = {}
+            self._persisted_fit_state_sidecar['predictive_datasets'] = {}
+            self.bayesian_result._set_has_distribution_cache(value=False)
+            self.bayesian_result._set_has_pair_cache(value=False)
+            self.bayesian_result._set_has_posterior_predictive(value=False)
+            return
+
+        flattened_samples = np.asarray(posterior_samples.flattened(), dtype=float)
+        parameter_names = list(posterior_samples.parameter_names)
+        if (
+            flattened_samples.ndim != _FLATTENED_POSTERIOR_SAMPLE_NDIM
+            or not parameter_names
+            or flattened_samples.shape[1] != len(parameter_names)
+        ):
+            self._persisted_fit_state_sidecar['distribution_caches'] = {}
+            self._persisted_fit_state_sidecar['pair_caches'] = {}
+            self._persisted_fit_state_sidecar['predictive_datasets'] = {}
+            self.bayesian_result._set_has_distribution_cache(value=False)
+            self.bayesian_result._set_has_pair_cache(value=False)
+            self.bayesian_result._set_has_posterior_predictive(value=False)
+            return
+
+        plotter = self.project.rendering.plotter
+        distribution_payload = self._store_bayesian_distribution_cache_projection(
+            plotter=plotter,
+            results=results,
+            flattened_samples=flattened_samples,
+            parameter_names=parameter_names,
+        )
+        pair_payload = self._store_bayesian_pair_cache_projection(
+            plotter=plotter,
+            results=results,
+            flattened_samples=flattened_samples,
+            parameter_names=parameter_names,
+        )
+        predictive_payload = self._store_bayesian_predictive_projection(
+            plotter=plotter,
+            results=results,
+        )
+
+        self._persisted_fit_state_sidecar['distribution_caches'] = distribution_payload
+        self._persisted_fit_state_sidecar['pair_caches'] = pair_payload
+        self._persisted_fit_state_sidecar['predictive_datasets'] = predictive_payload
+        self.bayesian_result._set_has_distribution_cache(value=bool(distribution_payload))
+        self.bayesian_result._set_has_pair_cache(value=bool(pair_payload))
+        self.bayesian_result._set_has_posterior_predictive(value=bool(predictive_payload))
+
+    def _store_bayesian_posterior_sidecar_projection(
+        self,
+        results: BayesianFitResults,
+    ) -> None:
+        """Persist posterior arrays while live samples exist."""
+        posterior_samples = results.posterior_samples
+        if posterior_samples is None:
+            self._persisted_fit_state_sidecar['posterior'] = {}
+            return
+
+        self._persisted_fit_state_sidecar['posterior'] = {
+            'parameter_samples': np.asarray(
+                posterior_samples.parameter_samples,
+                dtype=float,
+            ),
+            'log_posterior': (
+                None
+                if posterior_samples.log_posterior is None
+                else np.asarray(posterior_samples.log_posterior, dtype=float)
+            ),
+            'draw_index': (
+                None
+                if posterior_samples.draw_index is None
+                else np.asarray(posterior_samples.draw_index)
+            ),
+        }
+
+    def _store_bayesian_result_projection(self, results: BayesianFitResults) -> None:
+        """
+        Store Bayesian fit-result projections into persisted categories.
+        """
+        credible_interval_inner = 0.68
+        credible_interval_outer = 0.95
+        if len(results.credible_interval_levels) >= _CREDIBLE_INTERVAL_LEVEL_COUNT:
+            credible_interval_inner = float(results.credible_interval_levels[0])
+            credible_interval_outer = float(results.credible_interval_levels[1])
+
+        point_estimate_name = results.point_estimate_name or 'best_sample'
+        sampler_settings = results.sampler_settings
+        convergence = results.convergence_diagnostics
+
+        self.bayesian_result._set_sampler_name(results.sampler_name)
+        self.bayesian_result._set_point_estimate_name(point_estimate_name)
+        self.bayesian_result._set_success(value=results.success)
+        self.bayesian_result._set_sampler_completed(value=results.sampler_completed)
+        self.bayesian_result._set_best_log_posterior(results.best_log_posterior)
+        self.bayesian_result._set_credible_interval_inner(credible_interval_inner)
+        self.bayesian_result._set_credible_interval_outer(credible_interval_outer)
+        self.bayesian_result._set_has_posterior_samples(
+            value=results.posterior_samples is not None
+        )
+        self.bayesian_result._set_has_distribution_cache(value=False)
+        self.bayesian_result._set_has_pair_cache(value=False)
+        self.bayesian_result._set_has_posterior_predictive(value=False)
+        self.bayesian_result._set_sidecar_file('results.h5')
+        self._store_bayesian_posterior_sidecar_projection(results)
+
+        self.bayesian_sampler._set_steps(int(sampler_settings.get('steps', 0)))
+        self.bayesian_sampler._set_burn(int(sampler_settings.get('burn', 0)))
+        self.bayesian_sampler._set_thin(int(sampler_settings.get('thin', 0)))
+        self.bayesian_sampler._set_pop(int(sampler_settings.get('pop', 0)))
+        self.bayesian_sampler._set_parallel(int(sampler_settings.get('parallel', 0)))
+        self.bayesian_sampler._set_init(str(sampler_settings.get('init', '')))
+        random_seed = sampler_settings.get('random_seed')
+        self.bayesian_sampler._set_random_seed(None if random_seed is None else int(random_seed))
+
+        self.bayesian_convergence._set_converged(value=bool(convergence.get('converged', False)))
+        self.bayesian_convergence._set_max_r_hat(convergence.get('max_r_hat'))
+        self.bayesian_convergence._set_min_ess_bulk(convergence.get('min_ess_bulk'))
+        self.bayesian_convergence._set_n_draws(int(convergence.get('n_draws', 0)))
+        self.bayesian_convergence._set_n_chains(int(convergence.get('n_chains', 0)))
+        self.bayesian_convergence._set_n_parameters(int(convergence.get('n_parameters', 0)))
+
+        for summary in results.posterior_parameter_summaries:
+            self.bayesian_parameter_posteriors.create(summary=summary)
+
+        posterior_samples = results.posterior_samples
+        if posterior_samples is None:
+            return
+
+        self._store_bayesian_plot_cache_projection(results)
+        if len(posterior_samples.parameter_names) <= 1:
+            return
+
+        flattened = posterior_samples.flattened()
+        correlation_matrix = np.corrcoef(flattened, rowvar=False)
+        self._store_correlation_projection(
+            unique_names=list(posterior_samples.parameter_names),
+            correlation_matrix=correlation_matrix,
+            source_kind=FitCorrelationSourceEnum.POSTERIOR,
+        )
+
+    def _store_fit_result_projection(
+        self,
+        results: FitResults,
+        *,
+        experiments: list[object],
+        fitted_parameters: list[Parameter],
+    ) -> None:
+        """
+        Store the latest fit result into persisted fit-state categories.
+        """
+        if isinstance(results, BayesianFitResults):
+            self._store_common_fit_result_projection(
+                results,
+                result_kind=FitResultKindEnum.BAYESIAN,
+            )
+            self._store_bayesian_result_projection(results)
+            return
+
+        self._store_common_fit_result_projection(
+            results,
+            result_kind=FitResultKindEnum.DETERMINISTIC,
+        )
+        self._store_deterministic_result_projection(
+            results,
+            experiments=experiments,
+            fitted_parameters=fitted_parameters,
+        )
+
     def _resolve_sequential_data_dir(self) -> Path:
         """
         Resolve the sequential-fit data directory to an absolute path.
@@ -704,7 +1771,7 @@ def _resolve_sequential_data_dir(self) -> Path:
 
     def _prepare_fit_run(self) -> tuple[VerbosityEnum, object, object] | None:
         """Resolve common inputs for single and joint fitting."""
-        verb = VerbosityEnum(self.project.verbosity)
+        verb = VerbosityEnum(self.project.verbosity.fit.value)
         structures = self.project.structures
         if not structures:
             log.warning('No structures found in the project. Cannot run fit.')
@@ -718,6 +1785,7 @@ def _prepare_fit_run(self) -> tuple[VerbosityEnum, object, object] | None:
         # Apply constraints before fitting so that user-constrained
         # parameters are marked and excluded from the free parameter
         # list built by the fitter.
+        self._sync_live_minimizer_from_persisted_fit_state()
         self._update_categories()
 
         return verb, structures, experiments
@@ -768,6 +1836,7 @@ def _run_sequential(self) -> None:
 
         self._set_fitting_mode_type(FitModeEnum.SEQUENTIAL.value)
         self._update_categories()
+        self._clear_persisted_fit_state()
 
         max_workers_value = self._sequential_fit.max_workers.value
         max_workers = max_workers_value if max_workers_value == 'auto' else int(max_workers_value)
@@ -790,6 +1859,7 @@ def _run_sequential(self) -> None:
         finally:
             self.fit_results = None
             self.fitter.results = None
+            self._clear_persisted_fit_state()
 
         if self.project.info.path is not None:
             self.project.save()
diff --git a/src/easydiffraction/analysis/categories/__init__.py b/src/easydiffraction/analysis/categories/__init__.py
index 6c070cf1a..743e0ff02 100644
--- a/src/easydiffraction/analysis/categories/__init__.py
+++ b/src/easydiffraction/analysis/categories/__init__.py
@@ -3,8 +3,39 @@
 
 from easydiffraction.analysis.categories.aliases import Alias
 from easydiffraction.analysis.categories.aliases import Aliases
+from easydiffraction.analysis.categories.bayesian_convergence import BayesianConvergence
+from easydiffraction.analysis.categories.bayesian_distribution_caches import (
+    BayesianDistributionCacheItem,
+)
+from easydiffraction.analysis.categories.bayesian_distribution_caches import (
+    BayesianDistributionCaches,
+)
+from easydiffraction.analysis.categories.bayesian_pair_caches import BayesianPairCacheItem
+from easydiffraction.analysis.categories.bayesian_pair_caches import BayesianPairCaches
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors import (
+    BayesianParameterPosteriorItem,
+)
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors import (
+    BayesianParameterPosteriors,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets import (
+    BayesianPredictiveDatasetItem,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets import (
+    BayesianPredictiveDatasets,
+)
+from easydiffraction.analysis.categories.bayesian_result import BayesianResult
+from easydiffraction.analysis.categories.bayesian_sampler import BayesianSampler
 from easydiffraction.analysis.categories.constraints import Constraint
 from easydiffraction.analysis.categories.constraints import Constraints
+from easydiffraction.analysis.categories.deterministic_result import DeterministicResult
+from easydiffraction.analysis.categories.fit_parameter_correlations import (
+    FitParameterCorrelationItem,
+)
+from easydiffraction.analysis.categories.fit_parameter_correlations import FitParameterCorrelations
+from easydiffraction.analysis.categories.fit_parameters import FitParameterItem
+from easydiffraction.analysis.categories.fit_parameters import FitParameters
+from easydiffraction.analysis.categories.fit_result import FitResult
 from easydiffraction.analysis.categories.fitting import Fitting
 from easydiffraction.analysis.categories.joint_fit import JointFitCollection
 from easydiffraction.analysis.categories.joint_fit import JointFitItem
diff --git a/src/easydiffraction/analysis/categories/bayesian_convergence/__init__.py b/src/easydiffraction/analysis/categories/bayesian_convergence/__init__.py
new file mode 100644
index 000000000..e9acbd717
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_convergence/__init__.py
@@ -0,0 +1,7 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.bayesian_convergence.default import BayesianConvergence
+from easydiffraction.analysis.categories.bayesian_convergence.factory import (
+    BayesianConvergenceFactory,
+)
diff --git a/src/easydiffraction/analysis/categories/bayesian_convergence/default.py b/src/easydiffraction/analysis/categories/bayesian_convergence/default.py
new file mode 100644
index 000000000..48d00fa41
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_convergence/default.py
@@ -0,0 +1,121 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian convergence diagnostics category."""
+
+from __future__ import annotations
+
+from easydiffraction.analysis.categories.bayesian_convergence.factory import (
+    BayesianConvergenceFactory,
+)
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.variable import BoolDescriptor
+from easydiffraction.core.variable import IntegerDescriptor
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+@BayesianConvergenceFactory.register
+class BayesianConvergence(CategoryItem):
+    """Persisted Bayesian convergence diagnostics."""
+
+    _category_code = 'bayesian_convergence'
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted Bayesian convergence diagnostics',
+    )
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._converged = BoolDescriptor(
+            name='converged',
+            description='Whether the Bayesian fit met convergence criteria.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_bayesian_convergence.converged']),
+        )
+        self._max_r_hat = NumericDescriptor(
+            name='max_r_hat',
+            description='Maximum rank-normalized split-R-hat across parameters.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_convergence.max_r_hat']),
+        )
+        self._min_ess_bulk = NumericDescriptor(
+            name='min_ess_bulk',
+            description='Minimum bulk effective sample size across parameters.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_convergence.min_ess_bulk']),
+        )
+        self._n_draws = IntegerDescriptor(
+            name='n_draws',
+            description='Number of stored posterior draws.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_convergence.n_draws']),
+        )
+        self._n_chains = IntegerDescriptor(
+            name='n_chains',
+            description='Number of stored posterior chains.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_convergence.n_chains']),
+        )
+        self._n_parameters = IntegerDescriptor(
+            name='n_parameters',
+            description='Number of sampled parameters.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_convergence.n_parameters']),
+        )
+
+    @property
+    def converged(self) -> BoolDescriptor:
+        """Whether the Bayesian fit met convergence criteria."""
+        return self._converged
+
+    def _set_converged(self, *, value: bool) -> None:
+        """Set the convergence flag for internal callers."""
+        self._converged.value = value
+
+    @property
+    def max_r_hat(self) -> NumericDescriptor:
+        """Maximum rank-normalized split-R-hat across parameters."""
+        return self._max_r_hat
+
+    def _set_max_r_hat(self, value: float | None) -> None:
+        """Set the maximum R-hat for internal callers."""
+        self._max_r_hat.value = value
+
+    @property
+    def min_ess_bulk(self) -> NumericDescriptor:
+        """Minimum bulk effective sample size across parameters."""
+        return self._min_ess_bulk
+
+    def _set_min_ess_bulk(self, value: float | None) -> None:
+        """Set the minimum ESS bulk for internal callers."""
+        self._min_ess_bulk.value = value
+
+    @property
+    def n_draws(self) -> IntegerDescriptor:
+        """Number of stored posterior draws."""
+        return self._n_draws
+
+    def _set_n_draws(self, value: int) -> None:
+        """Set the draw count for internal callers."""
+        self._n_draws.value = value
+
+    @property
+    def n_chains(self) -> IntegerDescriptor:
+        """Number of stored posterior chains."""
+        return self._n_chains
+
+    def _set_n_chains(self, value: int) -> None:
+        """Set the chain count for internal callers."""
+        self._n_chains.value = value
+
+    @property
+    def n_parameters(self) -> IntegerDescriptor:
+        """Number of sampled parameters."""
+        return self._n_parameters
+
+    def _set_n_parameters(self, value: int) -> None:
+        """Set the sampled-parameter count for internal callers."""
+        self._n_parameters.value = value
diff --git a/src/easydiffraction/analysis/categories/bayesian_convergence/factory.py b/src/easydiffraction/analysis/categories/bayesian_convergence/factory.py
new file mode 100644
index 000000000..fbe5da383
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_convergence/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian-convergence factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class BayesianConvergenceFactory(FactoryBase):
+    """Create Bayesian-convergence categories by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/bayesian_distribution_caches/__init__.py b/src/easydiffraction/analysis/categories/bayesian_distribution_caches/__init__.py
new file mode 100644
index 000000000..4ecf63f0b
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_distribution_caches/__init__.py
@@ -0,0 +1,12 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.bayesian_distribution_caches.default import (
+    BayesianDistributionCacheItem,
+)
+from easydiffraction.analysis.categories.bayesian_distribution_caches.default import (
+    BayesianDistributionCaches,
+)
+from easydiffraction.analysis.categories.bayesian_distribution_caches.factory import (
+    BayesianDistributionCachesFactory,
+)
diff --git a/src/easydiffraction/analysis/categories/bayesian_distribution_caches/default.py b/src/easydiffraction/analysis/categories/bayesian_distribution_caches/default.py
new file mode 100644
index 000000000..41183dc39
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_distribution_caches/default.py
@@ -0,0 +1,151 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian distribution-cache manifest rows."""
+
+from __future__ import annotations
+
+from easydiffraction.analysis.categories.bayesian_distribution_caches.factory import (
+    BayesianDistributionCachesFactory,
+)
+from easydiffraction.core.category import CategoryCollection
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.validation import RegexValidator
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+class BayesianDistributionCacheItem(CategoryItem):
+    """Single persisted Bayesian distribution-cache manifest row."""
+
+    _category_code = 'bayesian_distribution_cache'
+    _category_entry_name = 'param_unique_name'
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._param_unique_name = StringDescriptor(
+            name='param_unique_name',
+            description='Unique parameter name for the cached distribution.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_.]*$'),
+            ),
+            cif_handler=CifHandler(names=['_bayesian_distribution_cache.param_unique_name']),
+        )
+        self._x_path = StringDescriptor(
+            name='x_path',
+            description='HDF5 dataset path for the distribution x-grid.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_distribution_cache.x_path']),
+        )
+        self._density_path = StringDescriptor(
+            name='density_path',
+            description='HDF5 dataset path for the cached density values.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_distribution_cache.density_path']),
+        )
+        self._n_grid = NumericDescriptor(
+            name='n_grid',
+            description='Number of grid points in the cached distribution.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_distribution_cache.n_grid']),
+        )
+        self._n_draws_cached = NumericDescriptor(
+            name='n_draws_cached',
+            description='Number of draws summarized into the cached distribution.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_distribution_cache.n_draws_cached']),
+        )
+
+    @property
+    def param_unique_name(self) -> StringDescriptor:
+        """Unique parameter name for the cached distribution."""
+        return self._param_unique_name
+
+    def _set_param_unique_name(self, value: str) -> None:
+        """Set the unique parameter name for internal callers."""
+        self._param_unique_name.value = value
+
+    @property
+    def x_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the distribution x-grid."""
+        return self._x_path
+
+    def _set_x_path(self, value: str) -> None:
+        """Set the x-grid dataset path for internal callers."""
+        self._x_path.value = value
+
+    @property
+    def density_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the cached density values."""
+        return self._density_path
+
+    def _set_density_path(self, value: str) -> None:
+        """Set the density dataset path for internal callers."""
+        self._density_path.value = value
+
+    @property
+    def n_grid(self) -> NumericDescriptor:
+        """Number of grid points in the cached distribution."""
+        return self._n_grid
+
+    def _set_n_grid(self, value: float) -> None:
+        """Set the grid-size count for internal callers."""
+        self._n_grid.value = value
+
+    @property
+    def n_draws_cached(self) -> NumericDescriptor:
+        """Number of draws summarized into the cached distribution."""
+        return self._n_draws_cached
+
+    def _set_n_draws_cached(self, value: float) -> None:
+        """Set the cached-draw count for internal callers."""
+        self._n_draws_cached.value = value
+
+
+@BayesianDistributionCachesFactory.register
+class BayesianDistributionCaches(CategoryCollection):
+    """Collection of persisted Bayesian distribution-cache manifests."""
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted Bayesian distribution-cache manifests',
+    )
+
+    def __init__(self) -> None:
+        super().__init__(item_type=BayesianDistributionCacheItem)
+
+    def create(
+        self,
+        *,
+        param_unique_name: str,
+        x_path: str,
+        density_path: str,
+        n_grid: float,
+        n_draws_cached: float,
+    ) -> None:
+        """
+        Create a persisted Bayesian distribution-cache manifest row.
+
+        Parameters
+        ----------
+        param_unique_name : str
+            Unique parameter name for the cached distribution.
+        x_path : str
+            HDF5 dataset path for the distribution x-grid.
+        density_path : str
+            HDF5 dataset path for the cached density values.
+        n_grid : float
+            Number of grid points in the cached distribution.
+        n_draws_cached : float
+            Number of draws summarized into the cached distribution.
+        """
+        item = BayesianDistributionCacheItem()
+        item._set_param_unique_name(param_unique_name)
+        item._set_x_path(x_path)
+        item._set_density_path(density_path)
+        item._set_n_grid(n_grid)
+        item._set_n_draws_cached(n_draws_cached)
+        self.add(item)
diff --git a/src/easydiffraction/analysis/categories/bayesian_distribution_caches/factory.py b/src/easydiffraction/analysis/categories/bayesian_distribution_caches/factory.py
new file mode 100644
index 000000000..a45016f5b
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_distribution_caches/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian-distribution-caches factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class BayesianDistributionCachesFactory(FactoryBase):
+    """Create Bayesian-distribution-cache collections by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/bayesian_pair_caches/__init__.py b/src/easydiffraction/analysis/categories/bayesian_pair_caches/__init__.py
new file mode 100644
index 000000000..edc955fa8
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_pair_caches/__init__.py
@@ -0,0 +1,8 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.bayesian_pair_caches.default import BayesianPairCacheItem
+from easydiffraction.analysis.categories.bayesian_pair_caches.default import BayesianPairCaches
+from easydiffraction.analysis.categories.bayesian_pair_caches.factory import (
+    BayesianPairCachesFactory,
+)
diff --git a/src/easydiffraction/analysis/categories/bayesian_pair_caches/default.py b/src/easydiffraction/analysis/categories/bayesian_pair_caches/default.py
new file mode 100644
index 000000000..937c77a5e
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_pair_caches/default.py
@@ -0,0 +1,267 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian pair-cache manifest rows."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+from easydiffraction.analysis.categories.bayesian_pair_caches.factory import (
+    BayesianPairCachesFactory,
+)
+from easydiffraction.core.category import CategoryCollection
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.validation import RegexValidator
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+def _normalized_parameter_pair(
+    param_unique_name_x: str,
+    param_unique_name_y: str,
+) -> tuple[str, str]:
+    """Return a stable ordering for a cached parameter pair."""
+    if param_unique_name_x <= param_unique_name_y:
+        return param_unique_name_x, param_unique_name_y
+    return param_unique_name_y, param_unique_name_x
+
+
+@dataclass(frozen=True, slots=True)
+class BayesianPairCachePaths:
+    """HDF5 dataset paths for one persisted pair cache."""
+
+    x_path: str
+    y_path: str
+    density_path: str
+    contour_level_path: str
+
+
+class BayesianPairCacheItem(CategoryItem):
+    """Single persisted Bayesian pair-cache manifest row."""
+
+    _category_code = 'bayesian_pair_cache'
+    _category_entry_name = 'id'
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._param_unique_name_x = StringDescriptor(
+            name='param_unique_name_x',
+            description='First unique parameter name in the cached pair.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_.]*$'),
+            ),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.param_unique_name_x']),
+        )
+        self._param_unique_name_y = StringDescriptor(
+            name='param_unique_name_y',
+            description='Second unique parameter name in the cached pair.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_.]*$'),
+            ),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.param_unique_name_y']),
+        )
+        self._id = StringDescriptor(
+            name='id',
+            description='Stable identifier for the cached parameter pair.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z0-9_.:-]+$'),
+            ),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.id']),
+        )
+        self._x_path = StringDescriptor(
+            name='x_path',
+            description='HDF5 dataset path for the pair-cache x-grid.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.x_path']),
+        )
+        self._y_path = StringDescriptor(
+            name='y_path',
+            description='HDF5 dataset path for the pair-cache y-grid.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.y_path']),
+        )
+        self._density_path = StringDescriptor(
+            name='density_path',
+            description='HDF5 dataset path for the pair-cache density grid.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.density_path']),
+        )
+        self._contour_level_path = StringDescriptor(
+            name='contour_level_path',
+            description='HDF5 dataset path for cached contour levels.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.contour_level_path']),
+        )
+        self._n_grid_x = NumericDescriptor(
+            name='n_grid_x',
+            description='Number of x-grid points in the cached pair.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.n_grid_x']),
+        )
+        self._n_grid_y = NumericDescriptor(
+            name='n_grid_y',
+            description='Number of y-grid points in the cached pair.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.n_grid_y']),
+        )
+        self._n_draws_cached = NumericDescriptor(
+            name='n_draws_cached',
+            description='Number of draws summarized into the cached pair.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_pair_cache.n_draws_cached']),
+        )
+
+    @property
+    def param_unique_name_x(self) -> StringDescriptor:
+        """First unique parameter name in the cached pair."""
+        return self._param_unique_name_x
+
+    def _set_param_unique_name_x(self, value: str) -> None:
+        """Set the first unique parameter name for internal callers."""
+        self._param_unique_name_x.value = value
+
+    @property
+    def param_unique_name_y(self) -> StringDescriptor:
+        """Second unique parameter name in the cached pair."""
+        return self._param_unique_name_y
+
+    def _set_param_unique_name_y(self, value: str) -> None:
+        """Set the second unique parameter name for internal callers."""
+        self._param_unique_name_y.value = value
+
+    @property
+    def id(self) -> StringDescriptor:
+        """Stable identifier for the cached parameter pair."""
+        return self._id
+
+    def _set_id(self, value: str) -> None:
+        """Set the pair-cache id for internal callers."""
+        self._id.value = value
+
+    @property
+    def x_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the pair-cache x-grid."""
+        return self._x_path
+
+    def _set_x_path(self, value: str) -> None:
+        """Set the pair-cache x-grid path for internal callers."""
+        self._x_path.value = value
+
+    @property
+    def y_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the pair-cache y-grid."""
+        return self._y_path
+
+    def _set_y_path(self, value: str) -> None:
+        """Set the pair-cache y-grid path for internal callers."""
+        self._y_path.value = value
+
+    @property
+    def density_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the pair-cache density grid."""
+        return self._density_path
+
+    def _set_density_path(self, value: str) -> None:
+        """Set the pair-cache density path for internal callers."""
+        self._density_path.value = value
+
+    @property
+    def contour_level_path(self) -> StringDescriptor:
+        """HDF5 dataset path for cached contour levels."""
+        return self._contour_level_path
+
+    def _set_contour_level_path(self, value: str) -> None:
+        """Set the contour-level path for internal callers."""
+        self._contour_level_path.value = value
+
+    @property
+    def n_grid_x(self) -> NumericDescriptor:
+        """Number of x-grid points in the cached pair."""
+        return self._n_grid_x
+
+    def _set_n_grid_x(self, value: float) -> None:
+        """Set the x-grid size for internal callers."""
+        self._n_grid_x.value = value
+
+    @property
+    def n_grid_y(self) -> NumericDescriptor:
+        """Number of y-grid points in the cached pair."""
+        return self._n_grid_y
+
+    def _set_n_grid_y(self, value: float) -> None:
+        """Set the y-grid size for internal callers."""
+        self._n_grid_y.value = value
+
+    @property
+    def n_draws_cached(self) -> NumericDescriptor:
+        """Number of draws summarized into the cached pair."""
+        return self._n_draws_cached
+
+    def _set_n_draws_cached(self, value: float) -> None:
+        """Set the cached-draw count for internal callers."""
+        self._n_draws_cached.value = value
+
+
+@BayesianPairCachesFactory.register
+class BayesianPairCaches(CategoryCollection):
+    """Collection of persisted Bayesian pair-cache manifests."""
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted Bayesian pair-cache manifests',
+    )
+
+    def __init__(self) -> None:
+        super().__init__(item_type=BayesianPairCacheItem)
+
+    def create(
+        self,
+        *,
+        parameter_names: tuple[str, str],
+        paths: BayesianPairCachePaths,
+        grid_shape: tuple[float, float],
+        n_draws_cached: float,
+        id: str | None = None,
+    ) -> None:
+        """
+        Create a persisted Bayesian pair-cache manifest row.
+
+        Parameters
+        ----------
+        parameter_names : tuple[str, str]
+            Unique parameter names for the cached pair.
+        paths : BayesianPairCachePaths
+            HDF5 dataset paths for the cached pair payloads.
+        grid_shape : tuple[float, float]
+            Number of x-grid and y-grid points in the cached pair.
+        n_draws_cached : float
+            Number of draws summarized into the cached pair.
+        id : str | None, default=None
+            Explicit persisted row id. When omitted, a simple sequential
+            identifier is generated.
+        """
+        param_unique_name_x, param_unique_name_y = parameter_names
+        normalized_x, normalized_y = _normalized_parameter_pair(
+            param_unique_name_x,
+            param_unique_name_y,
+        )
+        n_grid_x, n_grid_y = grid_shape
+        item = BayesianPairCacheItem()
+        item._set_param_unique_name_x(normalized_x)
+        item._set_param_unique_name_y(normalized_y)
+        item._set_x_path(paths.x_path)
+        item._set_y_path(paths.y_path)
+        item._set_density_path(paths.density_path)
+        item._set_contour_level_path(paths.contour_level_path)
+        item._set_n_grid_x(n_grid_x)
+        item._set_n_grid_y(n_grid_y)
+        item._set_n_draws_cached(n_draws_cached)
+        resolved_id = id or str(len(self) + 1)
+        item._set_id(resolved_id)
+        self.add(item)
diff --git a/src/easydiffraction/analysis/categories/bayesian_pair_caches/factory.py b/src/easydiffraction/analysis/categories/bayesian_pair_caches/factory.py
new file mode 100644
index 000000000..5c6df89b1
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_pair_caches/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian-pair-caches factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class BayesianPairCachesFactory(FactoryBase):
+    """Create Bayesian-pair-cache collections by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/__init__.py b/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/__init__.py
new file mode 100644
index 000000000..10ec76953
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/__init__.py
@@ -0,0 +1,12 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors.default import (
+    BayesianParameterPosteriorItem,
+)
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors.default import (
+    BayesianParameterPosteriors,
+)
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors.factory import (
+    BayesianParameterPosteriorsFactory,
+)
diff --git a/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/default.py b/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/default.py
new file mode 100644
index 000000000..ea234f0ec
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/default.py
@@ -0,0 +1,242 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian parameter posterior summary rows."""
+
+from __future__ import annotations
+
+from typing import TYPE_CHECKING
+
+from easydiffraction.analysis.categories.bayesian_parameter_posteriors.factory import (
+    BayesianParameterPosteriorsFactory,
+)
+from easydiffraction.core.category import CategoryCollection
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.validation import RegexValidator
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+if TYPE_CHECKING:
+    from easydiffraction.analysis.fit_helpers.bayesian import PosteriorParameterSummary
+
+
+class BayesianParameterPosteriorItem(CategoryItem):
+    """Single persisted Bayesian parameter posterior summary row."""
+
+    _category_code = 'bayesian_parameter_posterior'
+    _category_entry_name = 'unique_name'
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._unique_name = StringDescriptor(
+            name='unique_name',
+            description='Unique EasyDiffraction parameter name.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_.]*$'),
+            ),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.unique_name']),
+        )
+        self._display_name = StringDescriptor(
+            name='display_name',
+            description='Human-readable parameter label.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.display_name']),
+        )
+        self._best_sample_value = NumericDescriptor(
+            name='best_sample_value',
+            description='Committed sampled parameter value.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.best_sample_value']),
+        )
+        self._median = NumericDescriptor(
+            name='median',
+            description='Posterior median value.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.median']),
+        )
+        self._uncertainty = NumericDescriptor(
+            name='uncertainty',
+            description='Posterior standard deviation.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.uncertainty']),
+        )
+        self._interval_68_lower = NumericDescriptor(
+            name='interval_68_lower',
+            description='Lower bound of the 68% credible interval.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.interval_68_lower']),
+        )
+        self._interval_68_upper = NumericDescriptor(
+            name='interval_68_upper',
+            description='Upper bound of the 68% credible interval.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.interval_68_upper']),
+        )
+        self._interval_95_lower = NumericDescriptor(
+            name='interval_95_lower',
+            description='Lower bound of the 95% credible interval.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.interval_95_lower']),
+        )
+        self._interval_95_upper = NumericDescriptor(
+            name='interval_95_upper',
+            description='Upper bound of the 95% credible interval.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.interval_95_upper']),
+        )
+        self._ess_bulk = NumericDescriptor(
+            name='ess_bulk',
+            description='Bulk effective sample size when available.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.ess_bulk']),
+        )
+        self._r_hat = NumericDescriptor(
+            name='r_hat',
+            description='Rank-normalized split-R-hat when available.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_parameter_posterior.r_hat']),
+        )
+
+    @property
+    def unique_name(self) -> StringDescriptor:
+        """Unique EasyDiffraction parameter name."""
+        return self._unique_name
+
+    def _set_unique_name(self, value: str) -> None:
+        """Set the unique parameter name for internal callers."""
+        self._unique_name.value = value
+
+    @property
+    def display_name(self) -> StringDescriptor:
+        """Human-readable parameter label."""
+        return self._display_name
+
+    def _set_display_name(self, value: str) -> None:
+        """Set the display name for internal callers."""
+        self._display_name.value = value
+
+    @property
+    def best_sample_value(self) -> NumericDescriptor:
+        """Committed sampled parameter value."""
+        return self._best_sample_value
+
+    def _set_best_sample_value(self, value: float | None) -> None:
+        """Set the best sampled parameter value for internal callers."""
+        self._best_sample_value.value = value
+
+    @property
+    def median(self) -> NumericDescriptor:
+        """Posterior median value."""
+        return self._median
+
+    def _set_median(self, value: float | None) -> None:
+        """Set the posterior median for internal callers."""
+        self._median.value = value
+
+    @property
+    def uncertainty(self) -> NumericDescriptor:
+        """Posterior standard deviation."""
+        return self._uncertainty
+
+    def _set_uncertainty(self, value: float | None) -> None:
+        """Set the posterior uncertainty for internal callers."""
+        self._uncertainty.value = value
+
+    @property
+    def interval_68_lower(self) -> NumericDescriptor:
+        """Lower bound of the 68% credible interval."""
+        return self._interval_68_lower
+
+    def _set_interval_68_lower(self, value: float | None) -> None:
+        """Set the 68% interval lower bound for internal callers."""
+        self._interval_68_lower.value = value
+
+    @property
+    def interval_68_upper(self) -> NumericDescriptor:
+        """Upper bound of the 68% credible interval."""
+        return self._interval_68_upper
+
+    def _set_interval_68_upper(self, value: float | None) -> None:
+        """Set the 68% interval upper bound for internal callers."""
+        self._interval_68_upper.value = value
+
+    @property
+    def interval_95_lower(self) -> NumericDescriptor:
+        """Lower bound of the 95% credible interval."""
+        return self._interval_95_lower
+
+    def _set_interval_95_lower(self, value: float | None) -> None:
+        """Set the 95% interval lower bound for internal callers."""
+        self._interval_95_lower.value = value
+
+    @property
+    def interval_95_upper(self) -> NumericDescriptor:
+        """Upper bound of the 95% credible interval."""
+        return self._interval_95_upper
+
+    def _set_interval_95_upper(self, value: float | None) -> None:
+        """Set the 95% interval upper bound for internal callers."""
+        self._interval_95_upper.value = value
+
+    @property
+    def ess_bulk(self) -> NumericDescriptor:
+        """Bulk effective sample size when available."""
+        return self._ess_bulk
+
+    def _set_ess_bulk(self, value: float | None) -> None:
+        """Set the ESS bulk value for internal callers."""
+        self._ess_bulk.value = value
+
+    @property
+    def r_hat(self) -> NumericDescriptor:
+        """Rank-normalized split-R-hat when available."""
+        return self._r_hat
+
+    def _set_r_hat(self, value: float | None) -> None:
+        """Set the R-hat value for internal callers."""
+        self._r_hat.value = value
+
+
+@BayesianParameterPosteriorsFactory.register
+class BayesianParameterPosteriors(CategoryCollection):
+    """
+    Collection of persisted Bayesian parameter posterior summaries.
+    """
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted Bayesian parameter posterior summaries',
+    )
+
+    def __init__(self) -> None:
+        super().__init__(item_type=BayesianParameterPosteriorItem)
+
+    def create(
+        self,
+        *,
+        summary: PosteriorParameterSummary,
+    ) -> None:
+        """
+        Create a persisted Bayesian parameter posterior summary row.
+
+        Parameters
+        ----------
+        summary : PosteriorParameterSummary
+            Runtime posterior summary to persist.
+        """
+        item = BayesianParameterPosteriorItem()
+        item._set_unique_name(summary.unique_name)
+        item._set_display_name(summary.display_name)
+        item._set_best_sample_value(summary.best_sample_value)
+        item._set_median(summary.median)
+        item._set_uncertainty(summary.standard_deviation)
+        item._set_interval_68_lower(summary.interval_68[0])
+        item._set_interval_68_upper(summary.interval_68[1])
+        item._set_interval_95_lower(summary.interval_95[0])
+        item._set_interval_95_upper(summary.interval_95[1])
+        item._set_ess_bulk(summary.ess_bulk)
+        item._set_r_hat(summary.r_hat)
+        self.add(item)
diff --git a/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/factory.py b/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/factory.py
new file mode 100644
index 000000000..54eef2b83
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_parameter_posteriors/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian-parameter-posteriors factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class BayesianParameterPosteriorsFactory(FactoryBase):
+    """Create Bayesian-parameter-posterior collections by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/__init__.py b/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/__init__.py
new file mode 100644
index 000000000..f706de4fc
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/__init__.py
@@ -0,0 +1,12 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.bayesian_predictive_datasets.default import (
+    BayesianPredictiveDatasetItem,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets.default import (
+    BayesianPredictiveDatasets,
+)
+from easydiffraction.analysis.categories.bayesian_predictive_datasets.factory import (
+    BayesianPredictiveDatasetsFactory,
+)
diff --git a/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/default.py b/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/default.py
new file mode 100644
index 000000000..9bb8acf00
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/default.py
@@ -0,0 +1,260 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian predictive-dataset manifest rows."""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+from easydiffraction.analysis.categories.bayesian_predictive_datasets.factory import (
+    BayesianPredictiveDatasetsFactory,
+)
+from easydiffraction.core.category import CategoryCollection
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+@dataclass(frozen=True, slots=True)
+class BayesianPredictiveDatasetPaths:
+    """HDF5 dataset paths for one predictive dataset."""
+
+    x_path: str
+    best_sample_prediction_path: str
+    lower_95_path: str | None = None
+    upper_95_path: str | None = None
+    lower_68_path: str | None = None
+    upper_68_path: str | None = None
+    draws_path: str | None = None
+
+
+class BayesianPredictiveDatasetItem(CategoryItem):
+    """Single persisted Bayesian predictive-dataset manifest row."""
+
+    _category_code = 'bayesian_predictive_dataset'
+    _category_entry_name = 'experiment_name'
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._experiment_name = StringDescriptor(
+            name='experiment_name',
+            description='Experiment name for the cached predictive dataset.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.experiment_name']),
+        )
+        self._x_axis_name = StringDescriptor(
+            name='x_axis_name',
+            description='Name of the predictive dataset x-axis.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.x_axis_name']),
+        )
+        self._x_path = StringDescriptor(
+            name='x_path',
+            description='HDF5 dataset path for the predictive x-axis values.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.x_path']),
+        )
+        self._best_sample_prediction_path = StringDescriptor(
+            name='best_sample_prediction_path',
+            description='HDF5 dataset path for the committed predictive curve.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(
+                names=['_bayesian_predictive_dataset.best_sample_prediction_path']
+            ),
+        )
+        self._lower_95_path = StringDescriptor(
+            name='lower_95_path',
+            description='HDF5 dataset path for the lower 95% predictive band.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.lower_95_path']),
+        )
+        self._upper_95_path = StringDescriptor(
+            name='upper_95_path',
+            description='HDF5 dataset path for the upper 95% predictive band.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.upper_95_path']),
+        )
+        self._lower_68_path = StringDescriptor(
+            name='lower_68_path',
+            description='HDF5 dataset path for the lower 68% predictive band.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.lower_68_path']),
+        )
+        self._upper_68_path = StringDescriptor(
+            name='upper_68_path',
+            description='HDF5 dataset path for the upper 68% predictive band.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.upper_68_path']),
+        )
+        self._draws_path = StringDescriptor(
+            name='draws_path',
+            description='HDF5 dataset path for cached predictive draws.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.draws_path']),
+        )
+        self._n_x = NumericDescriptor(
+            name='n_x',
+            description='Number of x-axis points in the cached predictive dataset.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.n_x']),
+        )
+        self._n_draws_cached = NumericDescriptor(
+            name='n_draws_cached',
+            description='Number of cached predictive draws.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_predictive_dataset.n_draws_cached']),
+        )
+
+    @property
+    def experiment_name(self) -> StringDescriptor:
+        """Experiment name for the cached predictive dataset."""
+        return self._experiment_name
+
+    def _set_experiment_name(self, value: str) -> None:
+        """Set the experiment name for internal callers."""
+        self._experiment_name.value = value
+
+    @property
+    def x_axis_name(self) -> StringDescriptor:
+        """Name of the predictive dataset x-axis."""
+        return self._x_axis_name
+
+    def _set_x_axis_name(self, value: str) -> None:
+        """Set the x-axis name for internal callers."""
+        self._x_axis_name.value = value
+
+    @property
+    def x_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the predictive x-axis values."""
+        return self._x_path
+
+    def _set_x_path(self, value: str) -> None:
+        """Set the predictive x-axis path for internal callers."""
+        self._x_path.value = value
+
+    @property
+    def best_sample_prediction_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the committed predictive curve."""
+        return self._best_sample_prediction_path
+
+    def _set_best_sample_prediction_path(self, value: str) -> None:
+        """Set the best-sample prediction path for internal callers."""
+        self._best_sample_prediction_path.value = value
+
+    @property
+    def lower_95_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the lower 95% predictive band."""
+        return self._lower_95_path
+
+    def _set_lower_95_path(self, value: str | None) -> None:
+        """Set the lower-95 path for internal callers."""
+        self._lower_95_path.value = value
+
+    @property
+    def upper_95_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the upper 95% predictive band."""
+        return self._upper_95_path
+
+    def _set_upper_95_path(self, value: str | None) -> None:
+        """Set the upper-95 path for internal callers."""
+        self._upper_95_path.value = value
+
+    @property
+    def lower_68_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the lower 68% predictive band."""
+        return self._lower_68_path
+
+    def _set_lower_68_path(self, value: str | None) -> None:
+        """Set the lower-68 path for internal callers."""
+        self._lower_68_path.value = value
+
+    @property
+    def upper_68_path(self) -> StringDescriptor:
+        """HDF5 dataset path for the upper 68% predictive band."""
+        return self._upper_68_path
+
+    def _set_upper_68_path(self, value: str | None) -> None:
+        """Set the upper-68 path for internal callers."""
+        self._upper_68_path.value = value
+
+    @property
+    def draws_path(self) -> StringDescriptor:
+        """HDF5 dataset path for cached predictive draws."""
+        return self._draws_path
+
+    def _set_draws_path(self, value: str | None) -> None:
+        """Set the predictive-draws path for internal callers."""
+        self._draws_path.value = value
+
+    @property
+    def n_x(self) -> NumericDescriptor:
+        """Number of x-axis points in the cached predictive dataset."""
+        return self._n_x
+
+    def _set_n_x(self, value: float) -> None:
+        """Set the predictive x-axis size for internal callers."""
+        self._n_x.value = value
+
+    @property
+    def n_draws_cached(self) -> NumericDescriptor:
+        """Number of cached predictive draws."""
+        return self._n_draws_cached
+
+    def _set_n_draws_cached(self, value: float) -> None:
+        """Set the cached predictive-draw count for internal callers."""
+        self._n_draws_cached.value = value
+
+
+@BayesianPredictiveDatasetsFactory.register
+class BayesianPredictiveDatasets(CategoryCollection):
+    """Collection of persisted Bayesian predictive-dataset manifests."""
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted Bayesian predictive-dataset manifests',
+    )
+
+    def __init__(self) -> None:
+        super().__init__(item_type=BayesianPredictiveDatasetItem)
+
+    def create(
+        self,
+        *,
+        experiment_name: str,
+        x_axis_name: str,
+        paths: BayesianPredictiveDatasetPaths,
+        n_x: float,
+        n_draws_cached: float,
+    ) -> None:
+        """
+        Create a persisted Bayesian predictive-dataset manifest row.
+
+        Parameters
+        ----------
+        experiment_name : str
+            Experiment name for the cached predictive dataset.
+        x_axis_name : str
+            Name of the predictive dataset x-axis.
+        paths : BayesianPredictiveDatasetPaths
+            HDF5 dataset paths for the predictive dataset payloads.
+        n_x : float
+            Number of x-axis points in the cached predictive dataset.
+        n_draws_cached : float
+            Number of cached predictive draws.
+        """
+        item = BayesianPredictiveDatasetItem()
+        item._set_experiment_name(experiment_name)
+        item._set_x_axis_name(x_axis_name)
+        item._set_x_path(paths.x_path)
+        item._set_best_sample_prediction_path(paths.best_sample_prediction_path)
+        item._set_lower_95_path(paths.lower_95_path)
+        item._set_upper_95_path(paths.upper_95_path)
+        item._set_lower_68_path(paths.lower_68_path)
+        item._set_upper_68_path(paths.upper_68_path)
+        item._set_draws_path(paths.draws_path)
+        item._set_n_x(n_x)
+        item._set_n_draws_cached(n_draws_cached)
+        self.add(item)
diff --git a/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/factory.py b/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/factory.py
new file mode 100644
index 000000000..a88449263
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_predictive_datasets/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian-predictive-datasets factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class BayesianPredictiveDatasetsFactory(FactoryBase):
+    """Create Bayesian-predictive-dataset collections by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/bayesian_result/__init__.py b/src/easydiffraction/analysis/categories/bayesian_result/__init__.py
new file mode 100644
index 000000000..eab2a07f3
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_result/__init__.py
@@ -0,0 +1,5 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.bayesian_result.default import BayesianResult
+from easydiffraction.analysis.categories.bayesian_result.factory import BayesianResultFactory
diff --git a/src/easydiffraction/analysis/categories/bayesian_result/default.py b/src/easydiffraction/analysis/categories/bayesian_result/default.py
new file mode 100644
index 000000000..dd4210711
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_result/default.py
@@ -0,0 +1,215 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian fit-result metadata category."""
+
+from __future__ import annotations
+
+from easydiffraction.analysis.categories.bayesian_result.factory import BayesianResultFactory
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.variable import BoolDescriptor
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+@BayesianResultFactory.register
+class BayesianResult(CategoryItem):
+    """Persisted Bayesian fit-result metadata."""
+
+    _category_code = 'bayesian_result'
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted Bayesian fit-result metadata',
+    )
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._sampler_name = StringDescriptor(
+            name='sampler_name',
+            description='Name of the persisted Bayesian sampler.',
+            value_spec=AttributeSpec(default='dream'),
+            cif_handler=CifHandler(names=['_bayesian_result.sampler_name']),
+        )
+        self._point_estimate_name = StringDescriptor(
+            name='point_estimate_name',
+            description='Committed sampled point estimate name.',
+            value_spec=AttributeSpec(default='best_sample'),
+            cif_handler=CifHandler(names=['_bayesian_result.point_estimate_name']),
+        )
+        self._success = BoolDescriptor(
+            name='success',
+            description='Whether the persisted Bayesian fit produced usable results.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_bayesian_result.success']),
+        )
+        self._sampler_completed = BoolDescriptor(
+            name='sampler_completed',
+            description='Whether the sampler completed and returned posterior data.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_bayesian_result.sampler_completed']),
+        )
+        self._best_log_posterior = NumericDescriptor(
+            name='best_log_posterior',
+            description='Best log-posterior value reported by the sampler.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_result.best_log_posterior']),
+        )
+        self._credible_interval_inner = NumericDescriptor(
+            name='credible_interval_inner',
+            description='Inner credible-interval level used in summaries.',
+            value_spec=AttributeSpec(default=0.68),
+            cif_handler=CifHandler(names=['_bayesian_result.credible_interval_inner']),
+        )
+        self._credible_interval_outer = NumericDescriptor(
+            name='credible_interval_outer',
+            description='Outer credible-interval level used in summaries.',
+            value_spec=AttributeSpec(default=0.95),
+            cif_handler=CifHandler(names=['_bayesian_result.credible_interval_outer']),
+        )
+        self._has_posterior_samples = BoolDescriptor(
+            name='has_posterior_samples',
+            description='Whether posterior samples were persisted.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_bayesian_result.has_posterior_samples']),
+        )
+        self._has_distribution_cache = BoolDescriptor(
+            name='has_distribution_cache',
+            description='Whether distribution-cache manifests were persisted.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_bayesian_result.has_distribution_cache']),
+        )
+        self._has_pair_cache = BoolDescriptor(
+            name='has_pair_cache',
+            description='Whether pair-cache manifests were persisted.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_bayesian_result.has_pair_cache']),
+        )
+        self._has_posterior_predictive = BoolDescriptor(
+            name='has_posterior_predictive',
+            description='Whether posterior predictive manifests were persisted.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_bayesian_result.has_posterior_predictive']),
+        )
+        self._sidecar_file = StringDescriptor(
+            name='sidecar_file',
+            description='Relative path to the persisted Bayesian HDF5 sidecar.',
+            value_spec=AttributeSpec(default='results.h5'),
+            cif_handler=CifHandler(names=['_bayesian_result.sidecar_file']),
+        )
+
+    @property
+    def sampler_name(self) -> StringDescriptor:
+        """Name of the persisted Bayesian sampler."""
+        return self._sampler_name
+
+    def _set_sampler_name(self, value: str) -> None:
+        """Set the sampler name for internal callers."""
+        self._sampler_name.value = value
+
+    @property
+    def point_estimate_name(self) -> StringDescriptor:
+        """Committed sampled point estimate name."""
+        return self._point_estimate_name
+
+    def _set_point_estimate_name(self, value: str) -> None:
+        """Set the point-estimate name for internal callers."""
+        self._point_estimate_name.value = value
+
+    @property
+    def success(self) -> BoolDescriptor:
+        """
+        Whether the persisted Bayesian fit produced usable results.
+        """
+        return self._success
+
+    def _set_success(self, *, value: bool) -> None:
+        """Set the success flag for internal callers."""
+        self._success.value = value
+
+    @property
+    def sampler_completed(self) -> BoolDescriptor:
+        """Whether the sampler completed and returned posterior data."""
+        return self._sampler_completed
+
+    def _set_sampler_completed(self, *, value: bool) -> None:
+        """Set the sampler-completed flag for internal callers."""
+        self._sampler_completed.value = value
+
+    @property
+    def best_log_posterior(self) -> NumericDescriptor:
+        """Best log-posterior value reported by the sampler."""
+        return self._best_log_posterior
+
+    def _set_best_log_posterior(self, value: float | None) -> None:
+        """Set the best log-posterior for internal callers."""
+        self._best_log_posterior.value = value
+
+    @property
+    def credible_interval_inner(self) -> NumericDescriptor:
+        """Inner credible-interval level used in summaries."""
+        return self._credible_interval_inner
+
+    def _set_credible_interval_inner(self, value: float) -> None:
+        """
+        Set the inner credible-interval level for internal callers.
+        """
+        self._credible_interval_inner.value = value
+
+    @property
+    def credible_interval_outer(self) -> NumericDescriptor:
+        """Outer credible-interval level used in summaries."""
+        return self._credible_interval_outer
+
+    def _set_credible_interval_outer(self, value: float) -> None:
+        """
+        Set the outer credible-interval level for internal callers.
+        """
+        self._credible_interval_outer.value = value
+
+    @property
+    def has_posterior_samples(self) -> BoolDescriptor:
+        """Whether posterior samples were persisted."""
+        return self._has_posterior_samples
+
+    def _set_has_posterior_samples(self, *, value: bool) -> None:
+        """Set the posterior-samples flag for internal callers."""
+        self._has_posterior_samples.value = value
+
+    @property
+    def has_distribution_cache(self) -> BoolDescriptor:
+        """Whether distribution-cache manifests were persisted."""
+        return self._has_distribution_cache
+
+    def _set_has_distribution_cache(self, *, value: bool) -> None:
+        """Set the distribution-cache flag for internal callers."""
+        self._has_distribution_cache.value = value
+
+    @property
+    def has_pair_cache(self) -> BoolDescriptor:
+        """Whether pair-cache manifests were persisted."""
+        return self._has_pair_cache
+
+    def _set_has_pair_cache(self, *, value: bool) -> None:
+        """Set the pair-cache flag for internal callers."""
+        self._has_pair_cache.value = value
+
+    @property
+    def has_posterior_predictive(self) -> BoolDescriptor:
+        """Whether posterior predictive manifests were persisted."""
+        return self._has_posterior_predictive
+
+    def _set_has_posterior_predictive(self, *, value: bool) -> None:
+        """Set the posterior-predictive flag for internal callers."""
+        self._has_posterior_predictive.value = value
+
+    @property
+    def sidecar_file(self) -> StringDescriptor:
+        """Relative path to the persisted Bayesian HDF5 sidecar."""
+        return self._sidecar_file
+
+    def _set_sidecar_file(self, value: str) -> None:
+        """Set the sidecar-file path for internal callers."""
+        self._sidecar_file.value = value
diff --git a/src/easydiffraction/analysis/categories/bayesian_result/factory.py b/src/easydiffraction/analysis/categories/bayesian_result/factory.py
new file mode 100644
index 000000000..3d437a0df
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_result/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian-result factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class BayesianResultFactory(FactoryBase):
+    """Create Bayesian-result categories by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/bayesian_sampler/__init__.py b/src/easydiffraction/analysis/categories/bayesian_sampler/__init__.py
new file mode 100644
index 000000000..962e3fbae
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_sampler/__init__.py
@@ -0,0 +1,5 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.bayesian_sampler.default import BayesianSampler
+from easydiffraction.analysis.categories.bayesian_sampler.factory import BayesianSamplerFactory
diff --git a/src/easydiffraction/analysis/categories/bayesian_sampler/default.py b/src/easydiffraction/analysis/categories/bayesian_sampler/default.py
new file mode 100644
index 000000000..c75ffa156
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_sampler/default.py
@@ -0,0 +1,133 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Resolved Bayesian sampler settings category."""
+
+from __future__ import annotations
+
+from easydiffraction.analysis.categories.bayesian_sampler.factory import BayesianSamplerFactory
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.variable import IntegerDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+@BayesianSamplerFactory.register
+class BayesianSampler(CategoryItem):
+    """Persisted resolved Bayesian sampler settings."""
+
+    _category_code = 'bayesian_sampler'
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted resolved Bayesian sampler settings',
+    )
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._steps = IntegerDescriptor(
+            name='steps',
+            description='Resolved number of sampler steps.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_sampler.steps']),
+        )
+        self._burn = IntegerDescriptor(
+            name='burn',
+            description='Resolved burn-in count.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_sampler.burn']),
+        )
+        self._thin = IntegerDescriptor(
+            name='thin',
+            description='Resolved thinning interval.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_sampler.thin']),
+        )
+        self._pop = IntegerDescriptor(
+            name='pop',
+            description='Resolved population size.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_sampler.pop']),
+        )
+        self._parallel = IntegerDescriptor(
+            name='parallel',
+            description='Resolved DREAM worker count; 0 uses all CPUs.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_bayesian_sampler.parallel']),
+        )
+        self._init = StringDescriptor(
+            name='init',
+            description='Resolved DREAM initialization mode.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_bayesian_sampler.init']),
+        )
+        self._random_seed = IntegerDescriptor(
+            name='random_seed',
+            description='Resolved random seed used by the sampler.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_bayesian_sampler.random_seed']),
+        )
+
+    @property
+    def steps(self) -> IntegerDescriptor:
+        """Resolved number of sampler steps."""
+        return self._steps
+
+    def _set_steps(self, value: int) -> None:
+        """Set the step count for internal callers."""
+        self._steps.value = value
+
+    @property
+    def burn(self) -> IntegerDescriptor:
+        """Resolved burn-in count."""
+        return self._burn
+
+    def _set_burn(self, value: int) -> None:
+        """Set the burn-in count for internal callers."""
+        self._burn.value = value
+
+    @property
+    def thin(self) -> IntegerDescriptor:
+        """Resolved thinning interval."""
+        return self._thin
+
+    def _set_thin(self, value: int) -> None:
+        """Set the thinning interval for internal callers."""
+        self._thin.value = value
+
+    @property
+    def pop(self) -> IntegerDescriptor:
+        """Resolved population size."""
+        return self._pop
+
+    def _set_pop(self, value: int) -> None:
+        """Set the population size for internal callers."""
+        self._pop.value = value
+
+    @property
+    def parallel(self) -> IntegerDescriptor:
+        """Resolved DREAM worker count; 0 uses all CPUs."""
+        return self._parallel
+
+    def _set_parallel(self, value: int) -> None:
+        """Set the DREAM worker count for internal callers."""
+        self._parallel.value = value
+
+    @property
+    def init(self) -> StringDescriptor:
+        """Resolved DREAM initialization mode."""
+        return self._init
+
+    def _set_init(self, value: str) -> None:
+        """Set the initialization mode for internal callers."""
+        self._init.value = value
+
+    @property
+    def random_seed(self) -> IntegerDescriptor:
+        """Resolved random seed used by the sampler."""
+        return self._random_seed
+
+    def _set_random_seed(self, value: int | None) -> None:
+        """Set the random seed for internal callers."""
+        self._random_seed.value = value
diff --git a/src/easydiffraction/analysis/categories/bayesian_sampler/factory.py b/src/easydiffraction/analysis/categories/bayesian_sampler/factory.py
new file mode 100644
index 000000000..6f5d1033c
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/bayesian_sampler/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian-sampler factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class BayesianSamplerFactory(FactoryBase):
+    """Create Bayesian-sampler categories by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/constraints/default.py b/src/easydiffraction/analysis/categories/constraints/default.py
index d6f026221..0b622e183 100644
--- a/src/easydiffraction/analysis/categories/constraints/default.py
+++ b/src/easydiffraction/analysis/categories/constraints/default.py
@@ -137,7 +137,7 @@ def disable(self) -> None:
         """Deactivate constraints without deleting them."""
         self._enabled = False
 
-    def create(self, *, expression: str) -> None:
+    def create(self, *, expression: str, id: str | None = None) -> None:
         """
         Create a constraint from an expression string.
 
@@ -148,10 +148,15 @@ def create(self, *, expression: str) -> None:
         expression : str
             Constraint equation, e.g. ``'biso_Co2 = biso_Co1'`` or
             ``'occ_Ba = 1 - occ_La'``.
+        id : str | None, default=None
+            Explicit row identifier. When not ``None``, this value is
+            used as the collection key instead of the left-hand alias.
         """
         item = Constraint()
         item.expression = expression
-        if item.lhs_alias:
+        if id is not None:
+            item.id = id
+        elif item.lhs_alias:
             item.id = item.lhs_alias
         self.add(item)
         self._enabled = True
diff --git a/src/easydiffraction/analysis/categories/deterministic_result/__init__.py b/src/easydiffraction/analysis/categories/deterministic_result/__init__.py
new file mode 100644
index 000000000..18ed7af8f
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/deterministic_result/__init__.py
@@ -0,0 +1,7 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.deterministic_result.default import DeterministicResult
+from easydiffraction.analysis.categories.deterministic_result.factory import (
+    DeterministicResultFactory,
+)
diff --git a/src/easydiffraction/analysis/categories/deterministic_result/default.py b/src/easydiffraction/analysis/categories/deterministic_result/default.py
new file mode 100644
index 000000000..f66f2018d
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/deterministic_result/default.py
@@ -0,0 +1,187 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Deterministic fit-result metadata category."""
+
+from __future__ import annotations
+
+from easydiffraction.analysis.categories.deterministic_result.factory import (
+    DeterministicResultFactory,
+)
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.variable import BoolDescriptor
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+@DeterministicResultFactory.register
+class DeterministicResult(CategoryItem):
+    """Persisted deterministic fit-result metadata."""
+
+    _category_code = 'deterministic_result'
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted deterministic fit-result metadata',
+    )
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._optimizer_name = StringDescriptor(
+            name='optimizer_name',
+            description='Name of the persisted deterministic optimizer.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_deterministic_result.optimizer_name']),
+        )
+        self._method_name = StringDescriptor(
+            name='method_name',
+            description='Method name of the persisted deterministic optimizer.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_deterministic_result.method_name']),
+        )
+        self._objective_name = StringDescriptor(
+            name='objective_name',
+            description='Objective function name for the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_deterministic_result.objective_name']),
+        )
+        self._objective_value = NumericDescriptor(
+            name='objective_value',
+            description='Objective value for the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_deterministic_result.objective_value']),
+        )
+        self._n_data_points = NumericDescriptor(
+            name='n_data_points',
+            description='Number of data points used in the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_deterministic_result.n_data_points']),
+        )
+        self._n_parameters = NumericDescriptor(
+            name='n_parameters',
+            description='Number of parameters considered in the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_deterministic_result.n_parameters']),
+        )
+        self._n_free_parameters = NumericDescriptor(
+            name='n_free_parameters',
+            description='Number of free parameters in the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_deterministic_result.n_free_parameters']),
+        )
+        self._degrees_of_freedom = NumericDescriptor(
+            name='degrees_of_freedom',
+            description='Degrees of freedom for the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_deterministic_result.degrees_of_freedom']),
+        )
+        self._covariance_available = BoolDescriptor(
+            name='covariance_available',
+            description='Whether covariance was available for the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_deterministic_result.covariance_available']),
+        )
+        self._correlation_available = BoolDescriptor(
+            name='correlation_available',
+            description='Whether correlations were available for the persisted deterministic fit.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_deterministic_result.correlation_available']),
+        )
+
+    @property
+    def optimizer_name(self) -> StringDescriptor:
+        """Name of the persisted deterministic optimizer."""
+        return self._optimizer_name
+
+    def _set_optimizer_name(self, value: str) -> None:
+        """Set the optimizer name for internal callers."""
+        self._optimizer_name.value = value
+
+    @property
+    def method_name(self) -> StringDescriptor:
+        """Method name of the persisted deterministic optimizer."""
+        return self._method_name
+
+    def _set_method_name(self, value: str) -> None:
+        """Set the method name for internal callers."""
+        self._method_name.value = value
+
+    @property
+    def objective_name(self) -> StringDescriptor:
+        """
+        Objective function name for the persisted deterministic fit.
+        """
+        return self._objective_name
+
+    def _set_objective_name(self, value: str) -> None:
+        """Set the objective name for internal callers."""
+        self._objective_name.value = value
+
+    @property
+    def objective_value(self) -> NumericDescriptor:
+        """Objective value for the persisted deterministic fit."""
+        return self._objective_value
+
+    def _set_objective_value(self, value: float | None) -> None:
+        """Set the objective value for internal callers."""
+        self._objective_value.value = value
+
+    @property
+    def n_data_points(self) -> NumericDescriptor:
+        """
+        Number of data points used in the persisted deterministic fit.
+        """
+        return self._n_data_points
+
+    def _set_n_data_points(self, value: float) -> None:
+        """Set the data-point count for internal callers."""
+        self._n_data_points.value = value
+
+    @property
+    def n_parameters(self) -> NumericDescriptor:
+        """Number of parameters considered in the persisted fit."""
+        return self._n_parameters
+
+    def _set_n_parameters(self, value: float) -> None:
+        """Set the parameter count for internal callers."""
+        self._n_parameters.value = value
+
+    @property
+    def n_free_parameters(self) -> NumericDescriptor:
+        """
+        Number of free parameters in the persisted deterministic fit.
+        """
+        return self._n_free_parameters
+
+    def _set_n_free_parameters(self, value: float) -> None:
+        """Set the free-parameter count for internal callers."""
+        self._n_free_parameters.value = value
+
+    @property
+    def degrees_of_freedom(self) -> NumericDescriptor:
+        """Degrees of freedom for the persisted deterministic fit."""
+        return self._degrees_of_freedom
+
+    def _set_degrees_of_freedom(self, value: float) -> None:
+        """Set the degrees of freedom for internal callers."""
+        self._degrees_of_freedom.value = value
+
+    @property
+    def covariance_available(self) -> BoolDescriptor:
+        """Whether covariance was available for the persisted fit."""
+        return self._covariance_available
+
+    def _set_covariance_available(self, *, value: bool) -> None:
+        """Set the covariance-available flag for internal callers."""
+        self._covariance_available.value = value
+
+    @property
+    def correlation_available(self) -> BoolDescriptor:
+        """Whether correlations were available for the persisted fit."""
+        return self._correlation_available
+
+    def _set_correlation_available(self, *, value: bool) -> None:
+        """Set the correlation-available flag for internal callers."""
+        self._correlation_available.value = value
diff --git a/src/easydiffraction/analysis/categories/deterministic_result/factory.py b/src/easydiffraction/analysis/categories/deterministic_result/factory.py
new file mode 100644
index 000000000..1ee96acf9
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/deterministic_result/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Deterministic-result factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class DeterministicResultFactory(FactoryBase):
+    """Create deterministic-result categories by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/fit_parameter_correlations/__init__.py b/src/easydiffraction/analysis/categories/fit_parameter_correlations/__init__.py
new file mode 100644
index 000000000..bbb74736d
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_parameter_correlations/__init__.py
@@ -0,0 +1,12 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.fit_parameter_correlations.default import (
+    FitParameterCorrelationItem,
+)
+from easydiffraction.analysis.categories.fit_parameter_correlations.default import (
+    FitParameterCorrelations,
+)
+from easydiffraction.analysis.categories.fit_parameter_correlations.factory import (
+    FitParameterCorrelationsFactory,
+)
diff --git a/src/easydiffraction/analysis/categories/fit_parameter_correlations/default.py b/src/easydiffraction/analysis/categories/fit_parameter_correlations/default.py
new file mode 100644
index 000000000..4b08057e3
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_parameter_correlations/default.py
@@ -0,0 +1,184 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Persisted fit-parameter correlation summaries."""
+
+from __future__ import annotations
+
+from easydiffraction.analysis.categories.fit_parameter_correlations.factory import (
+    FitParameterCorrelationsFactory,
+)
+from easydiffraction.analysis.enums import FitCorrelationSourceEnum
+from easydiffraction.core.category import CategoryCollection
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.validation import MembershipValidator
+from easydiffraction.core.validation import RangeValidator
+from easydiffraction.core.validation import RegexValidator
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+def _normalized_parameter_pair(
+    param_unique_name_i: str,
+    param_unique_name_j: str,
+) -> tuple[str, str]:
+    """Return a stable ordering for a parameter pair."""
+    if param_unique_name_i <= param_unique_name_j:
+        return param_unique_name_i, param_unique_name_j
+    return param_unique_name_j, param_unique_name_i
+
+
+class FitParameterCorrelationItem(CategoryItem):
+    """Single persisted fit-parameter correlation row."""
+
+    _category_code = 'fit_parameter_correlation'
+    _category_entry_name = 'id'
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._id = StringDescriptor(
+            name='id',
+            description='Stable identifier for the persisted correlation row.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z0-9_.:-]+$'),
+            ),
+            cif_handler=CifHandler(names=['_fit_parameter_correlation.id']),
+        )
+        self._source_kind = StringDescriptor(
+            name='source_kind',
+            description='Origin of the persisted correlation summary.',
+            value_spec=AttributeSpec(
+                default=FitCorrelationSourceEnum.default().value,
+                validator=MembershipValidator(
+                    allowed=[member.value for member in FitCorrelationSourceEnum]
+                ),
+            ),
+            cif_handler=CifHandler(names=['_fit_parameter_correlation.source_kind']),
+        )
+        self._param_unique_name_i = StringDescriptor(
+            name='param_unique_name_i',
+            description='First unique parameter name in the persisted pair.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_.]*$'),
+            ),
+            cif_handler=CifHandler(names=['_fit_parameter_correlation.param_unique_name_i']),
+        )
+        self._param_unique_name_j = StringDescriptor(
+            name='param_unique_name_j',
+            description='Second unique parameter name in the persisted pair.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_.]*$'),
+            ),
+            cif_handler=CifHandler(names=['_fit_parameter_correlation.param_unique_name_j']),
+        )
+        self._correlation = NumericDescriptor(
+            name='correlation',
+            description='Persisted correlation coefficient for the parameter pair.',
+            value_spec=AttributeSpec(
+                default=0.0,
+                validator=RangeValidator(ge=-1.0, le=1.0),
+            ),
+            cif_handler=CifHandler(names=['_fit_parameter_correlation.correlation']),
+        )
+
+    @property
+    def id(self) -> StringDescriptor:
+        """Stable identifier for the persisted correlation row."""
+        return self._id
+
+    def _set_id(self, value: str) -> None:
+        """Set the persisted correlation-row id for internal callers."""
+        self._id.value = value
+
+    @property
+    def source_kind(self) -> StringDescriptor:
+        """Origin of the persisted correlation summary."""
+        return self._source_kind
+
+    def _set_source_kind(self, value: str) -> None:
+        """Set the correlation source kind for internal callers."""
+        self._source_kind.value = value
+
+    @property
+    def param_unique_name_i(self) -> StringDescriptor:
+        """First unique parameter name in the persisted pair."""
+        return self._param_unique_name_i
+
+    def _set_param_unique_name_i(self, value: str) -> None:
+        """Set the first parameter name for internal callers."""
+        self._param_unique_name_i.value = value
+
+    @property
+    def param_unique_name_j(self) -> StringDescriptor:
+        """Second unique parameter name in the persisted pair."""
+        return self._param_unique_name_j
+
+    def _set_param_unique_name_j(self, value: str) -> None:
+        """Set the second parameter name for internal callers."""
+        self._param_unique_name_j.value = value
+
+    @property
+    def correlation(self) -> NumericDescriptor:
+        """Persisted correlation coefficient for the parameter pair."""
+        return self._correlation
+
+    def _set_correlation(self, value: float) -> None:
+        """Set the correlation coefficient for internal callers."""
+        self._correlation.value = value
+
+
+@FitParameterCorrelationsFactory.register
+class FitParameterCorrelations(CategoryCollection):
+    """Collection of persisted fit-parameter correlation summaries."""
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted fit-parameter correlation summaries',
+    )
+
+    def __init__(self) -> None:
+        super().__init__(item_type=FitParameterCorrelationItem)
+
+    def create(
+        self,
+        *,
+        source_kind: str,
+        param_unique_name_i: str,
+        param_unique_name_j: str,
+        correlation: float,
+        id: str | None = None,
+    ) -> None:
+        """
+        Create a persisted fit-parameter correlation row.
+
+        Parameters
+        ----------
+        source_kind : str
+            Origin of the persisted correlation summary.
+        param_unique_name_i : str
+            First unique parameter name in the pair.
+        param_unique_name_j : str
+            Second unique parameter name in the pair.
+        correlation : float
+            Correlation coefficient for the parameter pair.
+        id : str | None, default=None
+            Explicit persisted row identifier. When omitted, a simple
+            sequential identifier is generated.
+        """
+        normalized_i, normalized_j = _normalized_parameter_pair(
+            param_unique_name_i,
+            param_unique_name_j,
+        )
+        item = FitParameterCorrelationItem()
+        item._set_source_kind(source_kind)
+        item._set_param_unique_name_i(normalized_i)
+        item._set_param_unique_name_j(normalized_j)
+        item._set_correlation(correlation)
+        resolved_id = id or str(len(self) + 1)
+        item._set_id(resolved_id)
+        self.add(item)
diff --git a/src/easydiffraction/analysis/categories/fit_parameter_correlations/factory.py b/src/easydiffraction/analysis/categories/fit_parameter_correlations/factory.py
new file mode 100644
index 000000000..095411086
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_parameter_correlations/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Fit-parameter-correlation factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class FitParameterCorrelationsFactory(FactoryBase):
+    """Create fit-parameter correlation collections by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/fit_parameters/__init__.py b/src/easydiffraction/analysis/categories/fit_parameters/__init__.py
new file mode 100644
index 000000000..64e72b416
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_parameters/__init__.py
@@ -0,0 +1,6 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.fit_parameters.default import FitParameterItem
+from easydiffraction.analysis.categories.fit_parameters.default import FitParameters
+from easydiffraction.analysis.categories.fit_parameters.factory import FitParametersFactory
diff --git a/src/easydiffraction/analysis/categories/fit_parameters/default.py b/src/easydiffraction/analysis/categories/fit_parameters/default.py
new file mode 100644
index 000000000..a5e1f94f5
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_parameters/default.py
@@ -0,0 +1,177 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Fit-parameter control snapshots."""
+
+from __future__ import annotations
+
+import numpy as np
+
+from easydiffraction.analysis.categories.fit_parameters.factory import FitParametersFactory
+from easydiffraction.core.category import CategoryCollection
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.validation import RegexValidator
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+class FitParameterItem(CategoryItem):
+    """Single persisted fit-parameter control row."""
+
+    _category_code = 'fit_parameter'
+    _category_entry_name = 'param_unique_name'
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._param_unique_name = StringDescriptor(
+            name='param_unique_name',
+            description='Unique name of the referenced live parameter.',
+            value_spec=AttributeSpec(
+                default='_',
+                validator=RegexValidator(pattern=r'^[A-Za-z_][A-Za-z0-9_.]*$'),
+            ),
+            cif_handler=CifHandler(names=['_fit_parameter.param_unique_name']),
+        )
+        self._fit_min = NumericDescriptor(
+            name='fit_min',
+            description='Persisted lower fit bound.',
+            value_spec=AttributeSpec(default=-np.inf),
+            cif_handler=CifHandler(names=['_fit_parameter.fit_min']),
+        )
+        self._fit_max = NumericDescriptor(
+            name='fit_max',
+            description='Persisted upper fit bound.',
+            value_spec=AttributeSpec(default=np.inf),
+            cif_handler=CifHandler(names=['_fit_parameter.fit_max']),
+        )
+        self._fit_bounds_uncertainty_multiplier = NumericDescriptor(
+            name='fit_bounds_uncertainty_multiplier',
+            description='Multiplier used to derive fit bounds from uncertainty.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_fit_parameter.fit_bounds_uncertainty_multiplier']),
+        )
+        self._start_value = NumericDescriptor(
+            name='start_value',
+            description='Persisted pre-fit value snapshot.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_fit_parameter.start_value']),
+        )
+        self._start_uncertainty = NumericDescriptor(
+            name='start_uncertainty',
+            description='Persisted pre-fit uncertainty snapshot.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_fit_parameter.start_uncertainty']),
+        )
+
+    @property
+    def param_unique_name(self) -> StringDescriptor:
+        """Unique name of the referenced live parameter."""
+        return self._param_unique_name
+
+    def _set_param_unique_name(self, value: str) -> None:
+        """
+        Set the referenced parameter unique name for internal callers.
+        """
+        self._param_unique_name.value = value
+
+    @property
+    def fit_min(self) -> NumericDescriptor:
+        """Persisted lower fit bound."""
+        return self._fit_min
+
+    def _set_fit_min(self, value: float) -> None:
+        """Set the persisted lower fit bound for internal callers."""
+        self._fit_min.value = value
+
+    @property
+    def fit_max(self) -> NumericDescriptor:
+        """Persisted upper fit bound."""
+        return self._fit_max
+
+    def _set_fit_max(self, value: float) -> None:
+        """Set the persisted upper fit bound for internal callers."""
+        self._fit_max.value = value
+
+    @property
+    def fit_bounds_uncertainty_multiplier(self) -> NumericDescriptor:
+        """Multiplier used to derive fit bounds from uncertainty."""
+        return self._fit_bounds_uncertainty_multiplier
+
+    def _set_fit_bounds_uncertainty_multiplier(
+        self,
+        value: float | None,
+    ) -> None:
+        """
+        Set the fit-bounds uncertainty multiplier for internal callers.
+        """
+        self._fit_bounds_uncertainty_multiplier.value = value
+
+    @property
+    def start_value(self) -> NumericDescriptor:
+        """Persisted pre-fit value snapshot."""
+        return self._start_value
+
+    def _set_start_value(self, value: float | None) -> None:
+        """Set the pre-fit value snapshot for internal callers."""
+        self._start_value.value = value
+
+    @property
+    def start_uncertainty(self) -> NumericDescriptor:
+        """Persisted pre-fit uncertainty snapshot."""
+        return self._start_uncertainty
+
+    def _set_start_uncertainty(self, value: float | None) -> None:
+        """Set the pre-fit uncertainty snapshot for internal callers."""
+        self._start_uncertainty.value = value
+
+
+@FitParametersFactory.register
+class FitParameters(CategoryCollection):
+    """Collection of persisted fit-parameter control snapshots."""
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Persisted fit-parameter control snapshots',
+    )
+
+    def __init__(self) -> None:
+        super().__init__(item_type=FitParameterItem)
+
+    def create(
+        self,
+        *,
+        param_unique_name: str,
+        fit_min: float,
+        fit_max: float,
+        fit_bounds_uncertainty_multiplier: float | None = None,
+        start_value: float | None = None,
+        start_uncertainty: float | None = None,
+    ) -> None:
+        """
+        Create a persisted fit-parameter control snapshot row.
+
+        Parameters
+        ----------
+        param_unique_name : str
+            Unique name of the referenced live parameter.
+        fit_min : float
+            Persisted lower fit bound.
+        fit_max : float
+            Persisted upper fit bound.
+        fit_bounds_uncertainty_multiplier : float | None, default=None
+            Multiplier used to derive fit bounds from uncertainty.
+        start_value : float | None, default=None
+            Persisted pre-fit value snapshot.
+        start_uncertainty : float | None, default=None
+            Persisted pre-fit uncertainty snapshot.
+        """
+        item = FitParameterItem()
+        item._set_param_unique_name(param_unique_name)
+        item._set_fit_min(fit_min)
+        item._set_fit_max(fit_max)
+        item._set_fit_bounds_uncertainty_multiplier(fit_bounds_uncertainty_multiplier)
+        item._set_start_value(start_value)
+        item._set_start_uncertainty(start_uncertainty)
+        self.add(item)
diff --git a/src/easydiffraction/analysis/categories/fit_parameters/factory.py b/src/easydiffraction/analysis/categories/fit_parameters/factory.py
new file mode 100644
index 000000000..4b73be48a
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_parameters/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Fit-parameter factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class FitParametersFactory(FactoryBase):
+    """Create fit-parameter collections by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/categories/fit_result/__init__.py b/src/easydiffraction/analysis/categories/fit_result/__init__.py
new file mode 100644
index 000000000..8578ab47b
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_result/__init__.py
@@ -0,0 +1,5 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from easydiffraction.analysis.categories.fit_result.default import FitResult
+from easydiffraction.analysis.categories.fit_result.factory import FitResultFactory
diff --git a/src/easydiffraction/analysis/categories/fit_result/default.py b/src/easydiffraction/analysis/categories/fit_result/default.py
new file mode 100644
index 000000000..564df008e
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_result/default.py
@@ -0,0 +1,135 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Common fit-result status category."""
+
+from __future__ import annotations
+
+from easydiffraction.analysis.categories.fit_result.factory import FitResultFactory
+from easydiffraction.analysis.enums import FitResultKindEnum
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.validation import MembershipValidator
+from easydiffraction.core.variable import BoolDescriptor
+from easydiffraction.core.variable import IntegerDescriptor
+from easydiffraction.core.variable import NumericDescriptor
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+
+
+@FitResultFactory.register
+class FitResult(CategoryItem):
+    """Common persisted fit-result status metadata."""
+
+    _category_code = 'fit_result'
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Common persisted fit-result status metadata',
+    )
+
+    def __init__(self) -> None:
+        super().__init__()
+        self._result_kind = StringDescriptor(
+            name='result_kind',
+            description='Kind of the latest persisted fit-result projection.',
+            value_spec=AttributeSpec(
+                default=FitResultKindEnum.default().value,
+                validator=MembershipValidator(
+                    allowed=[member.value for member in FitResultKindEnum]
+                ),
+            ),
+            cif_handler=CifHandler(names=['_fit_result.result_kind']),
+        )
+        self._success = BoolDescriptor(
+            name='success',
+            description='Whether the latest persisted fit-result projection succeeded.',
+            value_spec=AttributeSpec(default=False),
+            cif_handler=CifHandler(names=['_fit_result.success']),
+        )
+        self._message = StringDescriptor(
+            name='message',
+            description='Status message for the latest persisted fit-result projection.',
+            value_spec=AttributeSpec(default=''),
+            cif_handler=CifHandler(names=['_fit_result.message']),
+        )
+        self._iterations = IntegerDescriptor(
+            name='iterations',
+            description='Iteration count for the latest persisted fit-result projection.',
+            value_spec=AttributeSpec(default=0),
+            cif_handler=CifHandler(names=['_fit_result.iterations']),
+        )
+        self._fitting_time = NumericDescriptor(
+            name='fitting_time',
+            description='Fitting time in seconds for the latest persisted projection.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_fit_result.fitting_time']),
+        )
+        self._reduced_chi_square = NumericDescriptor(
+            name='reduced_chi_square',
+            description='Reduced chi-square for the latest persisted projection.',
+            value_spec=AttributeSpec(default=None, allow_none=True),
+            cif_handler=CifHandler(names=['_fit_result.reduced_chi_square']),
+        )
+
+    @property
+    def result_kind(self) -> StringDescriptor:
+        """Kind of the latest persisted fit-result projection."""
+        return self._result_kind
+
+    def _set_result_kind(self, value: str) -> None:
+        """Set the result kind for internal callers."""
+        self._result_kind.value = value
+
+    @property
+    def success(self) -> BoolDescriptor:
+        """
+        Whether the latest persisted fit-result projection succeeded.
+        """
+        return self._success
+
+    def _set_success(self, *, value: bool) -> None:
+        """Set the success flag for internal callers."""
+        self._success.value = value
+
+    @property
+    def message(self) -> StringDescriptor:
+        """
+        Status message for the latest persisted fit-result projection.
+        """
+        return self._message
+
+    def _set_message(self, value: str) -> None:
+        """Set the fit-result message for internal callers."""
+        self._message.value = value
+
+    @property
+    def iterations(self) -> IntegerDescriptor:
+        """
+        Iteration count for the latest persisted fit-result projection.
+        """
+        return self._iterations
+
+    def _set_iterations(self, value: int) -> None:
+        """Set the iteration count for internal callers."""
+        self._iterations.value = value
+
+    @property
+    def fitting_time(self) -> NumericDescriptor:
+        """
+        Fitting time in seconds for the latest persisted projection.
+        """
+        return self._fitting_time
+
+    def _set_fitting_time(self, value: float | None) -> None:
+        """Set the fitting time for internal callers."""
+        self._fitting_time.value = value
+
+    @property
+    def reduced_chi_square(self) -> NumericDescriptor:
+        """Reduced chi-square for the latest persisted projection."""
+        return self._reduced_chi_square
+
+    def _set_reduced_chi_square(self, value: float | None) -> None:
+        """Set the reduced chi-square for internal callers."""
+        self._reduced_chi_square.value = value
diff --git a/src/easydiffraction/analysis/categories/fit_result/factory.py b/src/easydiffraction/analysis/categories/fit_result/factory.py
new file mode 100644
index 000000000..66cbd32bc
--- /dev/null
+++ b/src/easydiffraction/analysis/categories/fit_result/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Fit-result factory."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class FitResultFactory(FactoryBase):
+    """Create fit-result categories by tag."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/analysis/enums.py b/src/easydiffraction/analysis/enums.py
index c1ef93f7b..94d78cc64 100644
--- a/src/easydiffraction/analysis/enums.py
+++ b/src/easydiffraction/analysis/enums.py
@@ -28,3 +28,27 @@ def description(self) -> str:
         if self is FitModeEnum.SEQUENTIAL:
             return 'Fit one experiment against a series of data files.'
         return ''
+
+
+class FitResultKindEnum(StrEnum):
+    """Persisted kind of the latest fit-result projection."""
+
+    DETERMINISTIC = 'deterministic'
+    BAYESIAN = 'bayesian'
+
+    @classmethod
+    def default(cls) -> FitResultKindEnum:
+        """Return the default persisted fit-result kind."""
+        return cls.DETERMINISTIC
+
+
+class FitCorrelationSourceEnum(StrEnum):
+    """Source of a persisted fit-parameter correlation summary."""
+
+    DETERMINISTIC = 'deterministic'
+    POSTERIOR = 'posterior'
+
+    @classmethod
+    def default(cls) -> FitCorrelationSourceEnum:
+        """Return the default persisted correlation source."""
+        return cls.DETERMINISTIC
diff --git a/src/easydiffraction/analysis/fit_helpers/tracking.py b/src/easydiffraction/analysis/fit_helpers/tracking.py
index 0f8e4c970..c24f94359 100644
--- a/src/easydiffraction/analysis/fit_helpers/tracking.py
+++ b/src/easydiffraction/analysis/fit_helpers/tracking.py
@@ -11,6 +11,8 @@
 from easydiffraction.analysis.fit_helpers.metrics import calculate_reduced_chi_square
 from easydiffraction.display.progress import ACTIVITY_LABEL_BURN_IN
 from easydiffraction.display.progress import ACTIVITY_LABEL_FITTING
+from easydiffraction.display.progress import ACTIVITY_LABEL_POST_PROCESSING
+from easydiffraction.display.progress import ACTIVITY_LABEL_PRE_PROCESSING
 from easydiffraction.display.progress import ACTIVITY_LABEL_PROCESSING
 from easydiffraction.display.progress import ACTIVITY_LABEL_SAMPLING
 from easydiffraction.display.progress import ActivityIndicator
@@ -28,6 +30,8 @@
 SAMPLER_PROGRESS_UPDATE_SECONDS = 5.0
 TRACKING_MODE_FIT = 'fit'
 TRACKING_MODE_SAMPLER = 'sampling'
+SAMPLER_PHASE_POST_PROCESSING = 'post-processing'
+SAMPLER_PHASE_PRE_PROCESSING = 'pre-processing'
 DEFAULT_HEADERS = ['iteration', 'time (s)', 'χ²', 'change / status']
 DEFAULT_ALIGNMENTS = ['center', 'center', 'center', 'center']
 SAMPLER_HEADERS = ['iteration', 'progress', 'time (s)', 'log posterior', 'phase']
@@ -85,6 +89,7 @@ def __init__(self) -> None:
         self._last_sampler_progress_percent: float | None = None
         self._last_sampler_log_posterior: float | None = None
         self._last_sampler_elapsed_time: float | None = None
+        self._sampler_pre_processing_pending: bool = False
 
         self._df_rows: list[list[str]] = []
         self._activity_indicator: ActivityIndicator | None = None
@@ -111,6 +116,7 @@ def reset(self) -> None:
         self._last_sampler_progress_percent = None
         self._last_sampler_log_posterior = None
         self._last_sampler_elapsed_time = None
+        self._sampler_pre_processing_pending = False
         self._df_rows = []
         self._activity_label = ACTIVITY_LABEL_FITTING
 
@@ -148,8 +154,25 @@ def track(
             self._last_chi2 = reduced_chi2
             return residuals
 
+        self.track_fit_progress(
+            iteration=self._iteration,
+            reduced_chi2=reduced_chi2,
+            elapsed_time=self._current_elapsed_time(),
+        )
+
+        return residuals
+
+    def track_fit_progress(
+        self,
+        *,
+        iteration: int,
+        reduced_chi2: float,
+        elapsed_time: float,
+    ) -> None:
+        """Update fit progress from a backend iteration callback."""
+        self._iteration = max(1, iteration)
+
         row: list[str] = []
-        elapsed_time = self._current_elapsed_time()
 
         if self._previous_chi2 is None:
             self._previous_chi2 = reduced_chi2
@@ -197,8 +220,6 @@ def track(
         self._last_chi2 = reduced_chi2
         self._last_iteration = self._iteration
 
-        return residuals
-
     def track_sampler_progress(self, update: SamplerProgressUpdate) -> None:
         """
         Update progress from a sampler monitor.
@@ -240,6 +261,37 @@ def track_sampler_progress(self, update: SamplerProgressUpdate) -> None:
         self._last_chi2 = update.reduced_chi2
         self._last_iteration = update.iteration
 
+    def start_sampler_pre_processing(self, *, total_iterations: int) -> None:
+        """Mark sampler setup so a status row appears on update."""
+        self._tracking_mode = TRACKING_MODE_SAMPLER
+        self._sampler_total_iterations = max(1, total_iterations)
+        self._last_sampler_phase = SAMPLER_PHASE_PRE_PROCESSING
+        self._last_sampler_progress_percent = None
+        self._last_sampler_log_posterior = None
+        self._last_sampler_elapsed_time = None
+        self._sampler_pre_processing_pending = True
+        self._set_activity_label(ACTIVITY_LABEL_PRE_PROCESSING)
+
+    def start_sampler_post_processing(
+        self,
+        *,
+        log_posterior: float | None = None,
+    ) -> None:
+        """Switch the activity indicator to post-processing."""
+        if self._tracking_mode != TRACKING_MODE_SAMPLER:
+            return
+
+        if self._sampler_total_iterations is None:
+            self._sampler_total_iterations = max(1, self._last_iteration or 1)
+
+        elapsed_time = self._elapsed_since_start()
+        self._last_sampler_phase = SAMPLER_PHASE_POST_PROCESSING
+        self._last_sampler_progress_percent = 100.0
+        if log_posterior is not None:
+            self._last_sampler_log_posterior = float(log_posterior)
+        self._last_sampler_elapsed_time = elapsed_time
+        self._set_activity_label(ACTIVITY_LABEL_POST_PROCESSING)
+
     @property
     def best_chi2(self) -> float | None:
         """Best recorded reduced chi-square value or None."""
@@ -273,6 +325,16 @@ def stop_timer(self) -> None:
         self._end_time = time.perf_counter()
         self._fitting_time = self._end_time - self._start_time
 
+    def _elapsed_since_start(self) -> float | None:
+        """
+        Return elapsed wall time using the active timer when available.
+        """
+        if self._start_time is None:
+            return None
+        if self._end_time is not None:
+            return self._end_time - self._start_time
+        return time.perf_counter() - self._start_time
+
     def start_tracking(self, minimizer_name: str, *, mode: str = TRACKING_MODE_FIT) -> None:
         """
         Initialize display and headers and announce the minimizer.
@@ -347,13 +409,21 @@ def _initial_sampler_progress_row(
         clamped_iteration: int,
         clamped_progress: float,
     ) -> list[str]:
-        if self._previous_chi2 is not None and self._best_chi2 is not None:
+        if self._df_rows:
             return []
 
         self._previous_chi2 = update.reduced_chi2
         self._best_chi2 = update.reduced_chi2
         self._best_iteration = update.iteration
         self._last_progress_time = update.elapsed_time
+        if self._sampler_pre_processing_pending:
+            self._sampler_pre_processing_pending = False
+            return self._sampler_status_row(
+                iteration_label=self._sampler_iteration_label(clamped_iteration),
+                phase=SAMPLER_PHASE_PRE_PROCESSING,
+                elapsed_time=update.elapsed_time,
+                log_posterior=update.log_posterior,
+            )
         return self._sampler_progress_row(
             clamped_iteration=clamped_iteration,
             clamped_progress=clamped_progress,
@@ -431,6 +501,25 @@ def _sampler_progress_row(
             phase,
         ]
 
+    def _sampler_status_row(
+        self,
+        *,
+        iteration_label: str = '',
+        phase: str,
+        elapsed_time: float | None,
+        log_posterior: float | None = None,
+    ) -> list[str]:
+        """
+        Return a status-only sampler row without iteration metrics.
+        """
+        return [
+            iteration_label,
+            '',
+            self._format_elapsed_time(elapsed_time),
+            '' if log_posterior is None else f'{log_posterior:.2f}',
+            phase,
+        ]
+
     def _finalize_sampler_tracking_row(self) -> None:
         row = self._final_sampler_tracking_row()
         if row is None:
@@ -453,6 +542,12 @@ def _final_sampler_tracking_row(self) -> list[str] | None:
 
         final_progress = self._resolved_final_sampler_progress()
         elapsed_time = self._resolved_final_sampler_elapsed_time()
+        if self._last_sampler_phase == SAMPLER_PHASE_POST_PROCESSING:
+            return self._sampler_status_row(
+                phase=SAMPLER_PHASE_POST_PROCESSING,
+                elapsed_time=elapsed_time,
+            )
+
         log_posterior = (
             f'{self._last_sampler_log_posterior:.2f}'
             if self._last_sampler_log_posterior is not None
@@ -602,8 +697,12 @@ def _default_activity_label(self) -> str:
     @staticmethod
     def _activity_label_for_sampler_phase(phase: str) -> str:
         normalized_phase = phase.strip().lower()
+        if normalized_phase == SAMPLER_PHASE_PRE_PROCESSING:
+            return ACTIVITY_LABEL_PRE_PROCESSING
         if normalized_phase == 'burn-in':
             return ACTIVITY_LABEL_BURN_IN
+        if normalized_phase == SAMPLER_PHASE_POST_PROCESSING:
+            return ACTIVITY_LABEL_POST_PROCESSING
         if normalized_phase == 'sampling':
             return ACTIVITY_LABEL_SAMPLING
         if normalized_phase:
diff --git a/src/easydiffraction/analysis/fitting.py b/src/easydiffraction/analysis/fitting.py
index 343c15742..72d89a6ce 100644
--- a/src/easydiffraction/analysis/fitting.py
+++ b/src/easydiffraction/analysis/fitting.py
@@ -21,6 +21,50 @@
     from easydiffraction.datablocks.structure.collection import Structures
 
 
+def _resolve_fit_result_message(results: FitResults) -> str:
+    """Return a normalized fit-result message."""
+    if results.message:
+        return results.message
+
+    raw_result = results.engine_result
+    message = getattr(raw_result, 'message', '')
+    return str(message) if message is not None else ''
+
+
+def _resolve_fit_result_iterations(results: FitResults) -> int:
+    """Return a normalized iteration or evaluation count."""
+    if results.iterations:
+        return int(results.iterations)
+
+    raw_result = results.engine_result
+    for attribute_name in ('nfev', 'nit', 'iterations', 'niter'):
+        value = getattr(raw_result, attribute_name, None)
+        if value is not None:
+            return int(value)
+    return 0
+
+
+def _resolve_fit_result_chi_square(results: FitResults) -> float | None:
+    """Return a normalized chi-square-like objective value."""
+    if results.chi_square is not None:
+        return float(results.chi_square)
+
+    raw_result = results.engine_result
+    chisqr = getattr(raw_result, 'chisqr', None)
+    if chisqr is not None:
+        return float(chisqr)
+
+    fun = getattr(raw_result, 'fun', None)
+    if fun is None:
+        return None
+
+    if np.isscalar(fun):
+        return float(fun)
+
+    fun_array = np.asarray(fun, dtype=float)
+    return float(np.sum(fun_array**2))
+
+
 class Fitter:
     """Handles the fitting workflow using a pluggable minimizer."""
 
@@ -30,6 +74,69 @@ def __init__(self, selection: str = MinimizerTypeEnum.default()) -> None:
         self.minimizer = MinimizerFactory.create(selection)
         self.results: FitResults | None = None
 
+    @staticmethod
+    def _collect_fit_parameters(
+        structures: Structures,
+        experiments: list[ExperimentBase],
+    ) -> list[Parameter]:
+        """Return free parameters from structures and experiments."""
+        expt_free_params: list[Parameter] = []
+        for expt in experiments:
+            expt_free_params.extend(
+                p
+                for p in expt.parameters
+                if isinstance(p, Parameter) and not p.user_constrained and p.free
+            )
+        return structures.free_parameters + expt_free_params
+
+    def _build_objective_function(
+        self,
+        *,
+        params: list[Parameter],
+        structures: Structures,
+        experiments: list[ExperimentBase],
+        weights: np.ndarray | None,
+        analysis: object,
+    ) -> object:
+        """Return the residual function for the current fit context."""
+
+        def objective_function(engine_params: dict[str, Any]) -> np.ndarray:
+            """Evaluate residuals for the current minimizer state."""
+            return self._residual_function(
+                engine_params=engine_params,
+                parameters=params,
+                structures=structures,
+                experiments=experiments,
+                weights=weights,
+                analysis=analysis,
+            )
+
+        return objective_function
+
+    def _postprocess_fit_results(
+        self,
+        *,
+        analysis: object,
+        experiments: list[ExperimentBase],
+        fitted_parameters: list[Parameter],
+    ) -> None:
+        """Populate result fields and persist fit projections."""
+        if self.results is None:
+            return
+
+        self.results.message = _resolve_fit_result_message(self.results)
+        self.results.iterations = _resolve_fit_result_iterations(self.results)
+        self.results.chi_square = _resolve_fit_result_chi_square(self.results)
+
+        if analysis is None:
+            return
+
+        analysis._store_fit_result_projection(
+            self.results,
+            experiments=experiments,
+            fitted_parameters=fitted_parameters,
+        )
+
     def fit(
         self,
         structures: Structures,
@@ -76,53 +183,48 @@ def fit(
             structure._need_categories_update = True
             structure._update_categories()
 
-        expt_free_params: list[Parameter] = []
-        for expt in experiments:
-            expt_free_params.extend(
-                p
-                for p in expt.parameters
-                if isinstance(p, Parameter) and not p.user_constrained and p.free
-            )
-        params = structures.free_parameters + expt_free_params
+        params = self._collect_fit_parameters(structures, experiments)
 
         if not params:
+            if analysis is not None:
+                analysis._clear_persisted_fit_state()
+                analysis.fit_results = None
+            self.results = None
             print('⚠️ No parameters selected for fitting.')
             return
 
+        if analysis is not None:
+            analysis._capture_fit_parameter_state(params)
+
         for param in params:
             param._fit_start_value = param.value
 
-        def objective_function(engine_params: dict[str, Any]) -> np.ndarray:
-            """
-            Evaluate the residual for the current minimizer parameters.
-
-            Parameters
-            ----------
-            engine_params : dict[str, Any]
-                Parameter values provided by the minimizer engine.
-
-            Returns
-            -------
-            np.ndarray
-                Residual array passed back to the minimizer.
-            """
-            return self._residual_function(
-                engine_params=engine_params,
-                parameters=params,
-                structures=structures,
-                experiments=experiments,
-                weights=weights,
+        objective_function = self._build_objective_function(
+            params=params,
+            structures=structures,
+            experiments=experiments,
+            weights=weights,
+            analysis=analysis,
+        )
+
+        try:
+            # Keep tracker finalization in this layer so post-processing
+            # can run before the live display is closed.
+            self.results = self.minimizer.fit(
+                params,
+                objective_function,
+                verbosity=verbosity,
+                finalize_tracking=False,
+                use_physical_limits=use_physical_limits,
+                random_seed=random_seed,
+            )
+            self._postprocess_fit_results(
                 analysis=analysis,
+                experiments=experiments,
+                fitted_parameters=params,
             )
-
-        # Perform fitting
-        self.results = self.minimizer.fit(
-            params,
-            objective_function,
-            verbosity=verbosity,
-            use_physical_limits=use_physical_limits,
-            random_seed=random_seed,
-        )
+        finally:
+            self.minimizer._stop_tracking()
 
     def _process_fit_results(
         self,
@@ -243,4 +345,7 @@ def _residual_function(
             # Append the residuals for this experiment
             residuals.extend(diff)
 
-        return self.minimizer.tracker.track(np.array(residuals), parameters)
+        residual_array = np.array(residuals)
+        if getattr(self.minimizer, '_tracks_progress_via_solver_monitor', lambda: False)():
+            return residual_array
+        return self.minimizer.tracker.track(residual_array, parameters)
diff --git a/src/easydiffraction/analysis/minimizers/base.py b/src/easydiffraction/analysis/minimizers/base.py
index bbc639f10..2d8bc83f9 100644
--- a/src/easydiffraction/analysis/minimizers/base.py
+++ b/src/easydiffraction/analysis/minimizers/base.py
@@ -34,7 +34,7 @@ def __init__(
     ) -> None:
         self.name: str | None = name
         self.method: str | None = method
-        self.max_iterations: int | None = max_iterations
+        self._max_iterations: int | None = max_iterations
         self.result: FitResults | None = None
         self._previous_chi2: float | None = None
         self._iteration: int | None = None
@@ -42,8 +42,19 @@ def __init__(
         self._best_iteration: int | None = None
         self._fitting_time: float | None = None
         self._resolved_random_seed: int | None = None
+        self._tracking_active: bool = False
+        self._deferred_warning_messages: list[str] = []
         self.tracker: FitProgressTracker = FitProgressTracker()
 
+    @property
+    def max_iterations(self) -> int | None:
+        """User-facing iteration limit for the current minimizer."""
+        return self._max_iterations
+
+    @max_iterations.setter
+    def max_iterations(self, value: int | None) -> None:
+        self._max_iterations = value
+
     def _start_tracking(
         self,
         minimizer_name: str,
@@ -61,19 +72,47 @@ def _start_tracking(
         """
         self.tracker.reset()
         self.tracker._verbosity = verbosity
+        self._tracking_active = True
+        self._deferred_warning_messages = []
         self.tracker.start_tracking(minimizer_name, mode=self._tracking_mode())
         self.tracker.start_timer()
 
     def _stop_tracking(self) -> None:
         """Stop timer and finalize tracking."""
+        if not self._tracking_active:
+            self._emit_deferred_warnings()
+            return
+
+        self._tracking_active = False
         self.tracker.stop_timer()
         self.tracker.finish_tracking()
+        if self.result is not None:
+            self.result.fitting_time = self.tracker.fitting_time
+        self._emit_deferred_warnings()
+
+    def _warn_after_tracking(self, message: str) -> None:
+        """Log immediately or defer a warning until tracking stops."""
+        if self._tracking_active:
+            self._deferred_warning_messages.append(message)
+            return
+
+        log.warning(message)
+
+    def _emit_deferred_warnings(self) -> None:
+        """Flush warnings deferred during live progress display."""
+        while self._deferred_warning_messages:
+            log.warning(self._deferred_warning_messages.pop(0))
 
     @staticmethod
     def _tracking_mode() -> str:
         """Return the tracker mode for the current minimizer."""
         return 'fit'
 
+    @staticmethod
+    def _tracks_progress_via_solver_monitor() -> bool:
+        """Return whether live progress comes from solver callbacks."""
+        return False
+
     @abstractmethod
     def _prepare_solver_args(self, parameters: list[Any]) -> dict[str, Any]:
         """
@@ -300,6 +339,7 @@ def fit(
         objective_function: Callable[..., object],
         verbosity: VerbosityEnum = VerbosityEnum.FULL,
         *,
+        finalize_tracking: bool = True,
         use_physical_limits: bool = False,
         random_seed: int | None = None,
     ) -> FitResults:
@@ -315,6 +355,8 @@ def fit(
             arguments.
         verbosity : VerbosityEnum, default=VerbosityEnum.FULL
             Console output verbosity.
+        finalize_tracking : bool, default=True
+            Whether to stop and finalize live tracking before returning.
         use_physical_limits : bool, default=False
             When ``True``, fall back to physical limits from the value
             spec for parameters whose ``fit_min``/``fit_max`` are
@@ -343,10 +385,10 @@ def fit(
             if resolved_random_seed is not None:
                 solver_args['random_seed'] = resolved_random_seed
             raw_result = self._run_solver(objective_function, **solver_args)
+            return self._finalize_fit(parameters, raw_result)
         finally:
-            self._stop_tracking()
-
-        return self._finalize_fit(parameters, raw_result)
+            if finalize_tracking:
+                self._stop_tracking()
 
     def _objective_function(
         self,
diff --git a/src/easydiffraction/analysis/minimizers/bumps.py b/src/easydiffraction/analysis/minimizers/bumps.py
index 109262343..f80446cfc 100644
--- a/src/easydiffraction/analysis/minimizers/bumps.py
+++ b/src/easydiffraction/analysis/minimizers/bumps.py
@@ -8,6 +8,7 @@
 from bumps.fitproblem import FitProblem
 from bumps.fitters import FITTERS
 from bumps.fitters import FitDriver
+from bumps.fitters import monitor as bumps_monitor
 from bumps.parameter import Parameter as BumpsParameter
 from scipy.optimize import OptimizeResult
 
@@ -20,38 +21,74 @@
 DEFAULT_MAX_ITERATIONS = 1000
 
 
+class _BumpsEvaluationLimitError(RuntimeError):
+    """Raised when the BUMPS residual-evaluation budget is exhausted."""
+
+    def __init__(
+        self,
+        *,
+        evaluation_count: int,
+        parameter_values: np.ndarray,
+        residuals: np.ndarray | None,
+    ) -> None:
+        super().__init__('maximum number of residual evaluations reached')
+        self.evaluation_count = evaluation_count
+        self.parameter_values = parameter_values
+        self.residuals = residuals
+
+
 class _EasyDiffractionFitness:
-    """
-    Adaptor wrapping an EasyDiffraction objective into bumps Fitness.
-    """
+    """Wrap an EasyDiffraction objective in the BUMPS fitness API."""
 
     def __init__(
         self,
         bumps_params: list[BumpsParameter],
         objective_function: object,
+        max_evaluations: int | None = None,
     ) -> None:
         self._bumps_params = bumps_params
         self._objective_function = objective_function
+        self._max_evaluations = max_evaluations
         self._numpoints = 0
+        self._evaluation_count = 0
+        self._count_evaluations = True
+        self._last_parameter_values: np.ndarray | None = None
+        self._last_residuals: np.ndarray | None = None
 
     def parameters(self) -> dict[str, BumpsParameter]:
-        """Return bumps parameters as a name-keyed dictionary."""
+        """Return BUMPS parameters as a name-keyed dictionary."""
         return {p.name: p for p in self._bumps_params}
 
     def update(self) -> None:
-        """Signal that parameters have changed (no-op)."""
+        """Signal that parameters have changed."""
 
     def residuals(self) -> np.ndarray:
-        """Compute residuals using current bumps parameter values."""
+        """Compute residuals for the current BUMPS parameter values."""
+        if (
+            self._count_evaluations
+            and self._max_evaluations is not None
+            and self._evaluation_count >= self._max_evaluations
+        ):
+            last_parameter_values = self._last_parameter_values
+            if last_parameter_values is None:
+                last_parameter_values = np.array([p.value for p in self._bumps_params])
+            raise _BumpsEvaluationLimitError(
+                evaluation_count=self._evaluation_count,
+                parameter_values=last_parameter_values,
+                residuals=self._last_residuals,
+            )
+
         values = np.array([p.value for p in self._bumps_params])
-        r = self._objective_function(values)
+        r = np.asarray(self._objective_function(values), dtype=float)
         self._numpoints = len(r)
+        self._last_parameter_values = values.copy()
+        self._last_residuals = r.copy()
+        if self._count_evaluations:
+            self._evaluation_count += 1
         return r
 
     def nllf(self) -> float:
-        """
-        Negative log-likelihood as half the sum of squared residuals.
-        """
+        """Return half the sum of squared residuals."""
         r = self.residuals()
         return 0.5 * np.sum(r**2)
 
@@ -59,10 +96,101 @@ def numpoints(self) -> int:
         """Return the number of data points."""
         return self._numpoints
 
+    @property
+    def evaluation_count(self) -> int:
+        """Return the live residual-evaluation count."""
+        return self._evaluation_count
+
+    def reset_evaluation_count(self) -> None:
+        """Reset the residual-evaluation counter."""
+        self._evaluation_count = 0
+
+    def stop_counting_evaluations(self) -> None:
+        """Freeze residual-evaluation counting."""
+        self._count_evaluations = False
+
+    @property
+    def last_residuals(self) -> np.ndarray | None:
+        """Return the last successful residual vector."""
+        return self._last_residuals
+
+    def last_reduced_chi_square(self, *, n_parameters: int) -> float | None:
+        """Return reduced chi-square for the last residual vector."""
+        if self._last_residuals is None:
+            return None
+
+        chi_square = float(np.sum(self._last_residuals**2))
+        dof = len(self._last_residuals) - n_parameters
+        if dof <= 0:
+            return chi_square
+        return chi_square / dof
+
+
+class _BumpsProgressMonitor(bumps_monitor.Monitor):
+    """Report live BUMPS fit evaluation counts."""
+
+    def __init__(
+        self,
+        *,
+        tracker: object,
+        fitness: _EasyDiffractionFitness,
+        n_points: int,
+        n_parameters: int,
+    ) -> None:
+        self._tracker = tracker
+        self._fitness = fitness
+        self._n_points = n_points
+        self._n_parameters = n_parameters
+
+    @staticmethod
+    def config_history(history: object) -> None:
+        """
+        Declare the history fields needed for deterministic progress.
+        """
+        history.requires(time=1, step=1, value=1)
+
+    def __call__(self, history: object) -> None:
+        """Forward deterministic BUMPS progress to the fit tracker."""
+        if not history.time or not history.value:
+            return
+
+        self._tracker.track_fit_progress(
+            iteration=self._reported_iteration(history),
+            reduced_chi2=self._reduced_chi_square_from_nllf(float(history.value[0])),
+            elapsed_time=float(history.time[0]),
+        )
+
+    def final(self, history: object, best: dict[str, object]) -> None:
+        """Record the final BUMPS state in the fit tracker."""
+        if not history.time or best.get('value') is None:
+            return
+
+        self._tracker.track_fit_progress(
+            iteration=self._reported_iteration(history),
+            reduced_chi2=self._reduced_chi_square_from_nllf(float(best['value'])),
+            elapsed_time=float(history.time[0]),
+        )
+
+    def _reported_iteration(self, history: object) -> int:
+        """Return the live fit evaluation count shown in progress."""
+        if self._fitness.evaluation_count > 0:
+            return self._fitness.evaluation_count
+
+        step = int(history.step[0]) if history.step else 0
+        return max(1, step)
+
+    def _reduced_chi_square_from_nllf(self, nllf: float) -> float:
+        """Convert negative log-likelihood to reduced chi-square."""
+        dof = self._n_points - self._n_parameters
+        chi_square = 2.0 * nllf
+        if dof <= 0:
+            return chi_square
+        return chi_square / dof
+
 
 @MinimizerFactory.register
 class BumpsMinimizer(MinimizerBase):
-    """Minimizer using the bumps package."""
+    """Minimizer using the BUMPS package."""
 
     type_info = TypeInfo(
         tag=MinimizerTypeEnum.BUMPS,
@@ -81,12 +209,19 @@ def __init__(
             max_iterations=max_iterations,
         )
 
+    @staticmethod
+    def _tracks_progress_via_solver_monitor() -> bool:
+        """
+        Use BUMPS monitor callbacks for live deterministic progress.
+        """
+        return True
+
     def _prepare_solver_args(  # noqa: PLR6301
         self,
         parameters: list[object],
     ) -> dict[str, object]:
         """
-        Prepare bumps parameters from EasyDiffraction parameters.
+        Prepare BUMPS parameters from EasyDiffraction parameters.
 
         Parameters
         ----------
@@ -116,7 +251,7 @@ def _run_solver(
         **kwargs: object,
     ) -> object:
         """
-        Run the bumps solver.
+        Run the BUMPS solver.
 
         Uses FitDriver directly instead of bumps.fitters.fit() to skip
         the expensive post-fit stderr/Jacobian computation that would
@@ -135,43 +270,123 @@ def _run_solver(
             A scipy OptimizeResult with the optimized values.
         """
         bumps_params = kwargs.get('bumps_params')
-        fitness = _EasyDiffractionFitness(bumps_params, objective_function)
+        fitness = _EasyDiffractionFitness(
+            bumps_params,
+            objective_function,
+            max_evaluations=self.max_iterations,
+        )
         fitness.nllf()  # pre-compute so numpoints() is valid
+        fitness.reset_evaluation_count()
         problem = FitProblem(fitness)
+        progress_monitor = _BumpsProgressMonitor(
+            tracker=self.tracker,
+            fitness=fitness,
+            n_points=fitness.numpoints(),
+            n_parameters=len(bumps_params),
+        )
 
         fitclass = next(cls for cls in FITTERS if cls.id == self.method)
         driver = FitDriver(
             fitclass=fitclass,
             problem=problem,
-            monitors=[],
+            monitors=[progress_monitor],
             steps=self.max_iterations,
         )
         driver.clip()
-        x, fx = driver.fit()
+        try:
+            x, fx = driver.fit()
+            evaluation_limit_reached = False
+            evaluation_limit_message = 'successful termination'
+        except _BumpsEvaluationLimitError as exc:
+            x, fx, evaluation_limit_message = self._handle_evaluation_limit(
+                exc=exc,
+                fitness=fitness,
+                n_parameters=len(bumps_params),
+            )
+            evaluation_limit_reached = True
+        finally:
+            fitness.stop_counting_evaluations()
 
-        success = x is not None
+        success = x is not None and not evaluation_limit_reached
         if success:
             problem.setp(x)
 
+        return self._build_optimize_result(
+            bumps_params=bumps_params,
+            fitness=fitness,
+            success=success,
+            evaluation_limit_reached=evaluation_limit_reached,
+            evaluation_limit_message=evaluation_limit_message,
+            function_value=fx,
+        )
+
+    def _handle_evaluation_limit(
+        self,
+        *,
+        exc: _BumpsEvaluationLimitError,
+        fitness: _EasyDiffractionFitness,
+        n_parameters: int,
+    ) -> tuple[np.ndarray, None, str]:
+        """
+        Build a partial result when the evaluation budget is exhausted.
+        """
+        reduced_chi2 = self._reduced_chi_square_from_limit(
+            exc=exc,
+            fitness=fitness,
+            n_parameters=n_parameters,
+        )
+        if reduced_chi2 is not None:
+            elapsed_time = self.tracker._current_elapsed_time()
+            self.tracker.track_fit_progress(
+                iteration=exc.evaluation_count,
+                reduced_chi2=reduced_chi2,
+                elapsed_time=0.0 if elapsed_time is None else elapsed_time,
+            )
+        return exc.parameter_values.copy(), None, str(exc)
+
+    @staticmethod
+    def _reduced_chi_square_from_limit(
+        *,
+        exc: _BumpsEvaluationLimitError,
+        fitness: _EasyDiffractionFitness,
+        n_parameters: int,
+    ) -> float | None:
+        """Return reduced chi-square at the evaluation cutoff."""
+        if exc.residuals is not None:
+            chi_square = float(np.sum(exc.residuals**2))
+            dof = len(exc.residuals) - n_parameters
+            return chi_square if dof <= 0 else chi_square / dof
+        if fitness.last_residuals is None:
+            return None
+        return fitness.last_reduced_chi_square(n_parameters=n_parameters)
+
+    def _build_optimize_result(
+        self,
+        *,
+        bumps_params: list[BumpsParameter],
+        fitness: _EasyDiffractionFitness,
+        success: bool,
+        evaluation_limit_reached: bool,
+        evaluation_limit_message: str,
+        function_value: float | None,
+    ) -> OptimizeResult:
+        """Convert the BUMPS solver outcome into an OptimizeResult."""
         # Read values back from bumps Parameters in our original order.
         # FitProblem sorts parameters alphabetically, so x from
         # driver.fit() uses that sorted order — not ours.
         result_x = np.array([p.value for p in bumps_params])
-
-        covar, stderr = (
+        covariance, stderr = (
             self._compute_covariance(bumps_params, fitness) if success else (None, None)
         )
-        var_names = [p.name for p in bumps_params]
-
         return OptimizeResult(
             x=result_x,
             dx=stderr,
-            fun=fx,
+            fun=function_value,
             success=success,
-            status=0 if success else -1,
-            message='successful termination' if success else 'fit failed',
-            covar=covar,
-            var_names=var_names,
+            status=0 if success else 5 if evaluation_limit_reached else -1,
+            message='successful termination' if success else evaluation_limit_message,
+            covar=covariance,
+            var_names=[p.name for p in bumps_params],
         )
 
     def _compute_covariance(  # noqa: PLR6301
@@ -225,7 +440,7 @@ def _sync_result_to_parameters(  # noqa: PLR6301
         raw_result: object,
     ) -> None:
         """
-        Synchronize the result from the solver to the parameters.
+        Synchronize the solver result back to the parameters.
 
         Parameters
         ----------
@@ -251,7 +466,7 @@ def _sync_result_to_parameters(  # noqa: PLR6301
 
     def _check_success(self, raw_result: object) -> bool:  # noqa: PLR6301
         """
-        Determine success from bumps OptimizeResult.
+        Determine success from a BUMPS OptimizeResult.
 
         Parameters
         ----------
diff --git a/src/easydiffraction/analysis/minimizers/bumps_dream.py b/src/easydiffraction/analysis/minimizers/bumps_dream.py
index 1cb42fb51..225c7ea0d 100644
--- a/src/easydiffraction/analysis/minimizers/bumps_dream.py
+++ b/src/easydiffraction/analysis/minimizers/bumps_dream.py
@@ -32,6 +32,8 @@
 from easydiffraction.core.metadata import TypeInfo
 from easydiffraction.utils.logging import log
 
+_BUMPS_DREAM_LOG = log
+
 DEFAULT_METHOD = 'dream'
 DEFAULT_MAX_ITERATIONS = 3000
 DEFAULT_BURN_FRACTION = 0.2
@@ -304,14 +306,28 @@ def __init__(
         self._parallel: int = DEFAULT_PARALLEL
         self._init: DreamPopulationInitializationEnum = DEFAULT_INIT
 
+    @property
+    def max_iterations(self) -> int:
+        """DREAM exposes sampler length through ``steps`` instead."""
+        sampler_name = self.type_info.description.partition('with ')[2].split()[0]
+        msg = f"{sampler_name} sampler uses 'steps' instead of 'max_iterations'."
+        raise AttributeError(msg)
+
+    @max_iterations.setter
+    def max_iterations(self, value: int) -> None:
+        del value
+        sampler_name = self.type_info.description.partition('with ')[2].split()[0]
+        msg = f"{sampler_name} sampler uses 'steps' instead of 'max_iterations'."
+        raise AttributeError(msg)
+
     @property
     def steps(self) -> int:
         """Number of DREAM generations retained after burn-in."""
-        return self._validated_positive_integer('steps', self.max_iterations)
+        return self._validated_positive_integer('steps', self._max_iterations)
 
     @steps.setter
     def steps(self, value: int) -> None:
-        self.max_iterations = self._validated_positive_integer('steps', value)
+        self._max_iterations = self._validated_positive_integer('steps', value)
 
     @property
     def burn(self) -> int | None:
@@ -601,6 +617,8 @@ def _run_solver(
         object
             Normalized DREAM result stored in an ``OptimizeResult``.
         """
+        total_iterations = int(self.steps + self._resolved_burn(self.steps) + 1)
+        self.tracker.start_sampler_pre_processing(total_iterations=total_iterations)
         context = self._prepare_run_context(objective_function=objective_function, kwargs=kwargs)
         driver_result = self._execute_driver(
             driver=context.driver,
@@ -621,6 +639,8 @@ def _run_solver(
                 sampler_completed=False,
             )
 
+        self.tracker.start_sampler_post_processing()
+
         return self._build_success_result(
             context=context,
             raw_state=driver_result.raw_state,
@@ -729,7 +749,7 @@ def _build_mapper(self, problem: FitProblem) -> object | None:
             return None
 
         if self._requires_serial_mapper_for_spawn_main_module():
-            log.warning(
+            self._warn_after_tracking(
                 'DREAM parallel evaluation requires an import-safe main '
                 'module on spawn-based multiprocessing; falling back to '
                 'serial execution.'
@@ -745,7 +765,7 @@ def _build_mapper(self, problem: FitProblem) -> object | None:
 
         try:
             if not can_pickle(problem):
-                log.warning(
+                self._warn_after_tracking(
                     'DREAM parallel evaluation requires a picklable '
                     'problem; falling back to serial execution.'
                 )
@@ -756,7 +776,7 @@ def _build_mapper(self, problem: FitProblem) -> object | None:
             message = str(error)
             if 'bootstrapping phase' not in message:
                 raise
-            log.warning(
+            self._warn_after_tracking(
                 'DREAM parallel evaluation requires an import-safe main '
                 'module on spawn-based multiprocessing; falling back to '
                 'serial execution.'
@@ -885,6 +905,10 @@ def _build_success_result(
             draw_index=np.asarray(draw_index, dtype=float),
         )
         convergence_diagnostics = compute_convergence_diagnostics(posterior_samples)
+        if not convergence_diagnostics.get('converged', True):
+            self._warn_after_tracking(
+                'Convergence diagnostics indicate the posterior may be poorly mixed.'
+            )
         posterior_parameter_summaries = summarize_posterior_parameters(
             parameter_names=context.parameter_names,
             posterior_samples=posterior_samples,
@@ -896,9 +920,6 @@ def _build_success_result(
             posterior_parameter_summaries
         )
 
-        if not convergence_diagnostics.get('converged', True):
-            log.warning('Convergence diagnostics indicate the posterior may be poorly mixed.')
-
         return OptimizeResult(
             x=best_sample_values,
             dx=posterior_standard_deviations,
diff --git a/src/easydiffraction/analysis/sequential.py b/src/easydiffraction/analysis/sequential.py
index 029486c79..a43b537a3 100644
--- a/src/easydiffraction/analysis/sequential.py
+++ b/src/easydiffraction/analysis/sequential.py
@@ -90,9 +90,9 @@ def _fit_worker(
     Returns
     -------
     dict[str, Any]
-        Result dict with keys: ``file_path``, ``fit_success``,
-        ``chi_squared``, ``reduced_chi_squared``, ``n_iterations``, and
-        per-parameter ``{unique_name}`` / ``{unique_name}.uncertainty``.
+        Result dict with keys: ``file_path``, ``success``,
+        ``reduced_chi_square``, ``iterations``, and per-parameter
+        ``{unique_name}`` / ``{unique_name}.uncertainty``.
     """
     # Lazy import to avoid circular dependencies and keep the module
     # importable without heavy imports at top level.
@@ -144,12 +144,12 @@ def _fit_worker(
         project.analysis.fitter = Fitter(template.minimizer_tag)
 
         # 10. Fit
-        original_verbosity = project.verbosity
-        project.verbosity = 'silent'
+        original_verbosity = project.verbosity.fit.value
+        project.verbosity.fit = 'silent'
         try:
             project.analysis.fit()
         finally:
-            project.verbosity = original_verbosity
+            project.verbosity.fit = original_verbosity
 
         # 11. Collect results
         result.update(_collect_results(project, template))
@@ -163,10 +163,9 @@ def _fit_worker(
         IndexError,
         OSError,
     ) as exc:
-        result['fit_success'] = False
-        result['chi_squared'] = None
-        result['reduced_chi_squared'] = None
-        result['n_iterations'] = 0
+        result['success'] = False
+        result['reduced_chi_square'] = None
+        result['iterations'] = 0
         result['error'] = str(exc)
 
     return result
@@ -352,17 +351,17 @@ def _collect_results(
 
     result: dict[str, Any] = {}
     fit_results = project.analysis.fit_results
+    tracker = getattr(project.analysis.fitter.minimizer, 'tracker', None)
+    best_iteration = getattr(tracker, 'best_iteration', None)
 
     if fit_results is not None:
-        result['fit_success'] = fit_results.success
-        result['chi_squared'] = fit_results.chi_square
-        result['reduced_chi_squared'] = fit_results.reduced_chi_square
-        result['n_iterations'] = project.analysis.fitter.minimizer.tracker.best_iteration or 0
+        result['fit_result.success'] = fit_results.success
+        result['fit_result.reduced_chi_square'] = fit_results.reduced_chi_square
+        result['fit_result.iterations'] = fit_results.iterations or best_iteration or 0
     else:
-        result['fit_success'] = False
-        result['chi_squared'] = None
-        result['reduced_chi_squared'] = None
-        result['n_iterations'] = 0
+        result['fit_result.success'] = False
+        result['fit_result.reduced_chi_square'] = None
+        result['fit_result.iterations'] = best_iteration or 0
 
     # Collect all free parameter values and uncertainties
     all_params = project.structures.parameters + project.experiments.parameters
@@ -383,10 +382,9 @@ def _collect_results(
 
 _META_COLUMNS = [
     'file_path',
-    'chi_squared',
-    'reduced_chi_squared',
-    'fit_success',
-    'n_iterations',
+    'fit_result.reduced_chi_square',
+    'fit_result.success',
+    'fit_result.iterations',
 ]
 
 
@@ -551,7 +549,8 @@ def _read_csv_for_recovery(
             file_path = row.get('file_path', '')
             if file_path:
                 fitted.add(_resolve_csv_file_path(csv_path, file_path))
-            if row.get('fit_success', '').lower() == 'true':
+            success_value = row.get('fit_result.success', row.get('success', ''))
+            if success_value.lower() == 'true':
                 params = _extract_params_from_row(row)
                 if params:
                     last_params = params
@@ -725,9 +724,9 @@ class SequentialRunPlan:
 def _summarize_chunk_results(results: list[dict[str, Any]]) -> tuple[str, str]:
     """Return average reduced chi-square and status for a chunk."""
     num_files = len(results)
-    successful = [r for r in results if r.get('fit_success')]
+    successful = [r for r in results if r.get('fit_result.success')]
     if successful:
-        avg_chi2 = sum(r['reduced_chi_squared'] for r in successful) / len(successful)
+        avg_chi2 = sum(r['fit_result.reduced_chi_square'] for r in successful) / len(successful)
         chi2_str = f'{avg_chi2:.2f}'
     else:
         chi2_str = '—'
@@ -798,10 +797,10 @@ def _build_file_progress_rows(
     rows: list[list[str]] = []
     time_str = _format_elapsed_seconds(elapsed_time)
     for index, result in enumerate(results, start=1):
-        reduced_chi2 = result.get('reduced_chi_squared')
+        reduced_chi2 = result.get('fit_result.reduced_chi_square')
         chi2_str = f'{reduced_chi2:.2f}' if reduced_chi2 is not None else '—'
-        iterations = str(result.get('n_iterations') or 0)
-        status = '✅' if result.get('fit_success') else '❌'
+        iterations = str(result.get('fit_result.iterations') or 0)
+        status = '✅' if result.get('fit_result.success') else '❌'
         rows.append([
             Path(result['file_path']).name,
             _format_progress_percent(completed_files_before + index, total_files),
@@ -897,7 +896,7 @@ def _print_sequential_completion(
         return
 
     console.print(f'✅ Sequential fitting complete: {processed_count} files processed.')
-    console.print(f'📄 Results saved to: {csv_path}')
+    console.print(f'📄 Results saved to:\n{csv_path}')
 
 
 def _prepare_sequential_run(
@@ -910,7 +909,7 @@ def _prepare_sequential_run(
     reverse: bool,
 ) -> SequentialRunPlan | None:
     """Resolve inputs and bookkeeping for one sequential-fit run."""
-    verbosity = VerbosityEnum(analysis.project.verbosity)
+    verbosity = VerbosityEnum(analysis.project.verbosity.fit.value)
 
     _check_seq_preconditions(analysis.project)
 
@@ -1260,7 +1259,7 @@ def _run_fit_loop(
 def _find_last_successful(results: list[dict[str, Any]]) -> dict[str, Any] | None:
     """Return the last successful result dict, or None."""
     for r in reversed(results):
-        if r.get('fit_success') and r.get('params'):
+        if r.get('fit_result.success') and r.get('params'):
             return r
     return None
 
diff --git a/src/easydiffraction/core/validation.py b/src/easydiffraction/core/validation.py
index 3fd882084..314d7e94c 100644
--- a/src/easydiffraction/core/validation.py
+++ b/src/easydiffraction/core/validation.py
@@ -27,6 +27,7 @@ class DataTypeHints:
     """Type hint aliases for numeric, string, and boolean types."""
 
     Numeric = int | float | np.integer | np.floating
+    Integer = int | np.integer
     String = str
     Bool = bool
 
@@ -38,6 +39,7 @@ class DataTypes(Enum):
     """Enumeration of supported data types for descriptors."""
 
     NUMERIC = (int, float, np.integer, np.floating)
+    INTEGER = (int, np.integer)
     STRING = (str,)
     BOOL = (bool,)
     ANY = (object,)  # fallback for unconstrained
diff --git a/src/easydiffraction/core/variable.py b/src/easydiffraction/core/variable.py
index 257a5771f..074afb714 100644
--- a/src/easydiffraction/core/variable.py
+++ b/src/easydiffraction/core/variable.py
@@ -276,6 +276,15 @@ def units(self) -> str:
 # ======================================================================
 
 
+class GenericIntegerDescriptor(GenericNumericDescriptor):
+    """Base descriptor that constrains values to integers."""
+
+    _value_type = DataTypes.INTEGER
+
+
+# ======================================================================
+
+
 class GenericParameter(GenericNumericDescriptor):
     """
     Numeric descriptor extended with fitting-related attributes.
@@ -461,6 +470,10 @@ def fit_bounds_uncertainty_multiplier(self) -> float | None:
         """
         return self._fit_bounds_uncertainty_multiplier
 
+    def _set_fit_bounds_uncertainty_multiplier(self, value: float | None) -> None:
+        """Set the cached uncertainty-derived fit-bounds multiplier."""
+        self._fit_bounds_uncertainty_multiplier = value
+
     def set_fit_bounds_from_uncertainty(
         self,
         multiplier: float = DEFAULT_FIT_BOUNDS_MULTIPLIER,
@@ -612,6 +625,33 @@ def __init__(
 # ======================================================================
 
 
+class IntegerDescriptor(GenericIntegerDescriptor):
+    """Integer descriptor bound to a CIF handler."""
+
+    def __init__(
+        self,
+        *,
+        cif_handler: CifHandler,
+        **kwargs: object,
+    ) -> None:
+        """
+        Integer descriptor bound to a CIF handler.
+
+        Parameters
+        ----------
+        cif_handler : CifHandler
+            Object that tracks CIF identifiers.
+        **kwargs : object
+            Forwarded to GenericIntegerDescriptor.
+        """
+        super().__init__(**kwargs)
+        self._cif_handler = cif_handler
+        self._cif_handler.attach(self)
+
+
+# ======================================================================
+
+
 class Parameter(GenericParameter):
     """Fittable parameter bound to a CIF handler."""
 
diff --git a/src/easydiffraction/datablocks/experiment/collection.py b/src/easydiffraction/datablocks/experiment/collection.py
index 16bfe5a49..65646d506 100644
--- a/src/easydiffraction/datablocks/experiment/collection.py
+++ b/src/easydiffraction/datablocks/experiment/collection.py
@@ -125,7 +125,7 @@ def add_from_data_path(
         scattering_type : str | None, default=None
             Scattering type (e.g. ``'bragg'``).
         """
-        verbosity = self._parent.verbosity if self._parent is not None else None
+        verbosity = self._parent.verbosity.fit.value if self._parent is not None else None
         verb = VerbosityEnum(verbosity) if verbosity is not None else VerbosityEnum.FULL
         experiment = ExperimentFactory.from_scratch(
             name=name,
diff --git a/src/easydiffraction/display/plotting.py b/src/easydiffraction/display/plotting.py
index 85cb1545a..3bfbd5ec5 100644
--- a/src/easydiffraction/display/plotting.py
+++ b/src/easydiffraction/display/plotting.py
@@ -16,6 +16,8 @@
 import numpy as np
 import pandas as pd
 
+from easydiffraction.analysis.enums import FitCorrelationSourceEnum
+from easydiffraction.analysis.enums import FitResultKindEnum
 from easydiffraction.analysis.fit_helpers.bayesian import PosteriorPredictiveSummary
 from easydiffraction.datablocks.experiment.item.base import intensity_category_for
 from easydiffraction.datablocks.experiment.item.enums import SampleFormEnum
@@ -89,6 +91,7 @@ class PosteriorPairPlotStyleEnum(StrEnum):
 POSTERIOR_FLATTENED_SAMPLE_NDIM = 2
 MIN_POSTERIOR_PARAMETER_COUNT = 2
 MIN_POSTERIOR_SAMPLE_COUNT = 2
+PAIR_DENSITY_SURFACE_NDIM = 2
 POSTERIOR_DENSITY_LINE_COLOR = 'rgb(99, 110, 250)'
 POSTERIOR_DENSITY_FILL_COLOR = 'rgba(99, 110, 250, 0.22)'
 POSTERIOR_PAIR_MARGINAL_DENSITY_LINE_COLOR = 'rgb(44, 160, 44)'
@@ -741,7 +744,7 @@ def plot_param_series(
         Parameters
         ----------
         param : object
-            Parameter descriptor whose ``unique_name`` identifies the
+            Descriptor whose ``unique_name`` or ``name`` identifies the
             values to plot.
         versus : str | None, default=None
             Persisted diffrn path (e.g.
@@ -749,7 +752,10 @@ def plot_param_series(
             column is used as the x-axis. When ``None``, the experiment
             sequence number is used instead.
         """
-        unique_name = param.unique_name
+        column_names = self._series_column_names(param)
+        if not column_names:
+            log.warning('Series plot target does not expose a CSV column name.')
+            return
 
         # Try CSV first (produced by fit_sequential or future fit)
         csv_path = None
@@ -761,19 +767,49 @@ def plot_param_series(
         if csv_path is not None:
             self._plot_param_series_from_csv(
                 csv_path=csv_path,
-                unique_name=unique_name,
+                column_names=column_names,
                 param_descriptor=param,
                 versus_path=versus,
             )
         else:
             # Fallback: in-memory snapshots from fit() single mode
             self.plot_param_series_from_snapshots(
-                unique_name,
+                column_names[0],
                 versus,
                 self._project.experiments,
                 self._project.analysis._parameter_snapshots,
             )
 
+    @staticmethod
+    def _series_column_names(param: object) -> list[str]:
+        """Return candidate CSV column names for one plotted series."""
+        names: list[str] = []
+
+        unique_name = getattr(param, 'unique_name', None)
+        if isinstance(unique_name, str) and unique_name:
+            names.append(unique_name)
+
+        name = getattr(param, 'name', None)
+        if isinstance(name, str) and name and name not in names:
+            names.append(name)
+
+        return names
+
+    @staticmethod
+    def _numeric_series_values(values: object) -> list[float]:
+        """Return one CSV column normalized to numeric plot values."""
+        series = pd.Series(values)
+        if series.dtype == bool:
+            return series.astype(float).tolist()
+
+        normalized = series.replace({
+            'True': 1.0,
+            'False': 0.0,
+            'true': 1.0,
+            'false': 0.0,
+        })
+        return pd.to_numeric(normalized, errors='raise').tolist()
+
     def plot_all_param_series(
         self,
         versus: str | None = None,
@@ -1573,13 +1609,15 @@ def _get_param_correlation_dataframe(self) -> pd.DataFrame | None:
             return corr_df
 
         raw_result = self._raw_fit_result_for_correlation(fit_results)
-        if raw_result is None:
-            return None
+        if raw_result is not None:
+            corr_df = self._correlation_dataframe_from_engine_result(
+                raw_result=raw_result,
+                parameters=fit_results.parameters,
+            )
+            if corr_df is not None:
+                return corr_df
 
-        corr_df = self._correlation_dataframe_from_engine_result(
-            raw_result=raw_result,
-            parameters=fit_results.parameters,
-        )
+        corr_df = self._correlation_dataframe_from_persisted_projection(fit_results)
         if corr_df is not None:
             return corr_df
 
@@ -1606,15 +1644,71 @@ def _raw_fit_result_for_correlation(fit_results: object) -> object | None:
         if raw_result is None:
             raw_result = getattr(fit_results, 'engine_result', None)
         if raw_result is None:
-            log.warning('No raw fit result available. Correlation matrix cannot be plotted.')
             return None
 
         var_names = getattr(raw_result, 'var_names', None)
         if not var_names:
-            log.warning('Fit result does not expose variable names for a correlation matrix.')
             return None
         return raw_result
 
+    def _correlation_dataframe_from_persisted_projection(
+        self,
+        fit_results: object,
+    ) -> pd.DataFrame | None:
+        """
+        Return correlations restored from persisted fit-state rows.
+        """
+        if self._project is None:
+            return None
+
+        analysis = self._project.analysis
+        source_kind = (
+            FitCorrelationSourceEnum.POSTERIOR.value
+            if analysis.fit_result.result_kind.value == FitResultKindEnum.BAYESIAN.value
+            else FitCorrelationSourceEnum.DETERMINISTIC.value
+        )
+        correlation_rows = [
+            row
+            for row in analysis.fit_parameter_correlations
+            if row.source_kind.value == source_kind
+        ]
+        if not correlation_rows:
+            return None
+
+        parameter_names = [
+            getattr(parameter, 'unique_name', '')
+            for parameter in getattr(fit_results, 'parameters', [])
+            if getattr(parameter, 'unique_name', None)
+        ]
+        if not parameter_names:
+            parameter_names = [
+                summary.unique_name
+                for summary in getattr(fit_results, 'posterior_parameter_summaries', [])
+            ]
+
+        for row in correlation_rows:
+            parameter_names.extend([row.param_unique_name_i.value, row.param_unique_name_j.value])
+        parameter_names = list(dict.fromkeys(parameter_names))
+        if len(parameter_names) < MIN_POSTERIOR_PARAMETER_COUNT:
+            return None
+
+        correlation_values = np.eye(len(parameter_names), dtype=float)
+        corr_df = pd.DataFrame(
+            correlation_values,
+            index=parameter_names,
+            columns=parameter_names,
+        )
+        wrote_any = False
+        for row in correlation_rows:
+            i_name = row.param_unique_name_i.value
+            j_name = row.param_unique_name_j.value
+            if i_name not in corr_df.index or j_name not in corr_df.index:
+                continue
+            corr_df.loc[i_name, j_name] = float(row.correlation.value)
+            corr_df.loc[j_name, i_name] = float(row.correlation.value)
+            wrote_any = True
+        return corr_df if wrote_any else None
+
     def _correlation_dataframe_from_engine_result(
         self,
         *,
@@ -2431,7 +2525,7 @@ def _plot_legend_background_color(self) -> str:
             return legend_background_color()
         return PlotlyPlotter._legend_background_color()
 
-    def _posterior_contour_traces(
+    def _resolved_posterior_contour_surface(
         self,
         *,
         fit_results: object,
@@ -2440,11 +2534,14 @@ def _posterior_contour_traces(
         x_values: np.ndarray,
         y_values: np.ndarray,
         grid_size: int,
-    ) -> tuple[object, object] | None:
-        """
-        Return filled and line contour traces for posterior pair plots.
-        """
-        go = __import__('plotly.graph_objects', fromlist=['Contour'])
+    ) -> tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray | None] | None:
+        """Return cached or computed posterior contour surface data."""
+        cached_surface = self._cached_posterior_pair_surface(
+            x_parameter_name=x_parameter_name,
+            y_parameter_name=y_parameter_name,
+        )
+        if cached_surface is not None:
+            return cached_surface
 
         bounds = self._posterior_pair_bounds(
             fit_results=fit_results,
@@ -2464,13 +2561,44 @@ def _posterior_contour_traces(
             return None
 
         x_grid, y_grid, density = density_surface
+        return x_grid, y_grid, density, None
+
+    def _posterior_contour_traces(
+        self,
+        *,
+        fit_results: object,
+        x_parameter_name: str,
+        y_parameter_name: str,
+        x_values: np.ndarray,
+        y_values: np.ndarray,
+        grid_size: int,
+    ) -> tuple[object, object] | None:
+        """
+        Return filled and line contour traces for posterior pair plots.
+        """
+        go = __import__('plotly.graph_objects', fromlist=['Contour'])
+
+        surface = self._resolved_posterior_contour_surface(
+            fit_results=fit_results,
+            x_parameter_name=x_parameter_name,
+            y_parameter_name=y_parameter_name,
+            x_values=x_values,
+            y_values=y_values,
+            grid_size=grid_size,
+        )
+        if surface is None:
+            return None
+
+        x_grid, y_grid, density, contour_levels = surface
+
         fill_colorscale, line_colorscale = self._posterior_pair_contour_colorscales(
             x_values,
             y_values,
         )
-        contour_start = float(np.max(density) * 0.20)
-        contour_end = float(np.max(density) * 0.95)
-        contour_size = float(np.max(density) * 0.15)
+        contour_start, contour_end, contour_size = self._posterior_contour_levels(
+            density=density,
+            contour_levels=contour_levels,
+        )
         fill_density = np.array(density, copy=True)
         fill_density[fill_density < contour_start] = np.nan
         fill_trace = go.Contour(
@@ -2516,6 +2644,80 @@ def _posterior_contour_traces(
         )
         return fill_trace, line_trace
 
+    def _cached_posterior_pair_surface(
+        self,
+        *,
+        x_parameter_name: str,
+        y_parameter_name: str,
+    ) -> tuple[np.ndarray, np.ndarray, np.ndarray, np.ndarray | None] | None:
+        """
+        Return a restored posterior pair-density surface when available.
+        """
+        if self._project is None:
+            return None
+
+        analysis = self._project.analysis
+        sidecar_data = getattr(analysis, '_persisted_fit_state_sidecar', {})
+        pair_caches = sidecar_data.get('pair_caches', {})
+        for cache in analysis.bayesian_pair_caches:
+            cache_x = cache.param_unique_name_x.value
+            cache_y = cache.param_unique_name_y.value
+            if {cache_x, cache_y} != {x_parameter_name, y_parameter_name}:
+                continue
+
+            cache_data = pair_caches.get(cache.id.value)
+            if cache_data is None:
+                return None
+
+            x_grid = np.asarray(cache_data.get('x'), dtype=float)
+            y_grid = np.asarray(cache_data.get('y'), dtype=float)
+            density = np.asarray(cache_data.get('density'), dtype=float)
+            contour_levels = cache_data.get('contour_levels')
+            if contour_levels is not None:
+                contour_levels = np.asarray(contour_levels, dtype=float)
+
+            if x_parameter_name != cache_x or y_parameter_name != cache_y:
+                x_grid, y_grid = y_grid, x_grid
+                if density.ndim == PAIR_DENSITY_SURFACE_NDIM:
+                    density = density.T
+
+            expected_shape = (y_grid.size, x_grid.size)
+            if x_grid.ndim != 1 or y_grid.ndim != 1 or density.shape != expected_shape:
+                log.warning(
+                    'Persisted posterior pair cache is invalid for '
+                    f'{x_parameter_name!r} and {y_parameter_name!r}.'
+                )
+                return None
+            return x_grid, y_grid, density, contour_levels
+
+        return None
+
+    @staticmethod
+    def _posterior_contour_levels(
+        *,
+        density: np.ndarray,
+        contour_levels: np.ndarray | None,
+    ) -> tuple[float, float, float]:
+        """Return contour start, end, and step for pair density."""
+        if contour_levels is not None and contour_levels.ndim == 1 and contour_levels.size > 0:
+            finite_levels = contour_levels[np.isfinite(contour_levels)]
+            if finite_levels.size > 0:
+                start = float(finite_levels[0])
+                end = float(finite_levels[-1])
+                if finite_levels.size > 1:
+                    size = float(np.min(np.diff(finite_levels)))
+                else:
+                    size = max(end - start, abs(end) * 0.15, 1e-6)
+                if end > start and size > 0:
+                    return start, end, size
+
+        density_max = float(np.max(density))
+        return (
+            density_max * 0.20,
+            density_max * 0.95,
+            density_max * 0.15,
+        )
+
     def _build_param_distribution_plot(
         self,
         param: object,
@@ -2608,11 +2810,13 @@ def _plot_ascii_param_distribution(
             fit_results=context.fit_results,
             parameter_name=context.parameter_name,
         )
-        density_curve = self._posterior_density_curve(
-            context.values,
-            lower_bound=lower_bound,
-            upper_bound=upper_bound,
-        )
+        density_curve = self._cached_posterior_density_curve(context.parameter_name)
+        if density_curve is None:
+            density_curve = self._posterior_density_curve(
+                context.values,
+                lower_bound=lower_bound,
+                upper_bound=upper_bound,
+            )
         if density_curve is None:
             log.warning(
                 f'Posterior distribution is unavailable for parameter {context.parameter_name}.'
@@ -2629,6 +2833,30 @@ def _plot_ascii_param_distribution(
             height=self.height,
         )
 
+    def _cached_posterior_density_curve(
+        self,
+        parameter_name: str,
+    ) -> tuple[np.ndarray, np.ndarray] | None:
+        """
+        Return a restored posterior density curve for one parameter.
+        """
+        if self._project is None:
+            return None
+
+        sidecar_data = getattr(self._project.analysis, '_persisted_fit_state_sidecar', {})
+        cache_data = sidecar_data.get('distribution_caches', {}).get(parameter_name)
+        if cache_data is None:
+            return None
+
+        x_values = np.asarray(cache_data.get('x'), dtype=float)
+        density_values = np.asarray(cache_data.get('density'), dtype=float)
+        if x_values.ndim != 1 or density_values.shape != x_values.shape:
+            log.warning(
+                f'Persisted posterior distribution cache is invalid for {parameter_name!r}.'
+            )
+            return None
+        return x_values, density_values
+
     def _posterior_distribution_context(
         self,
         param: object,
@@ -2974,11 +3202,13 @@ def _posterior_density_trace(
             fit_results=fit_results,
             parameter_name=parameter_name,
         )
-        density_curve = self._posterior_density_curve(
-            values,
-            lower_bound=lower_bound,
-            upper_bound=upper_bound,
-        )
+        density_curve = self._cached_posterior_density_curve(parameter_name)
+        if density_curve is None:
+            density_curve = self._posterior_density_curve(
+                values,
+                lower_bound=lower_bound,
+                upper_bound=upper_bound,
+            )
         if density_curve is None:
             return None
 
@@ -3312,8 +3542,7 @@ def _get_or_build_posterior_predictive_summary(
             return None
 
         posterior_predictive = getattr(fit_results, 'posterior_predictive', None)
-        posterior_samples = getattr(fit_results, 'posterior_samples', None)
-        if posterior_predictive is None or posterior_samples is None:
+        if posterior_predictive is None:
             return None
 
         x_axis_name = getattr(x_axis, 'value', x_axis)
@@ -3331,9 +3560,18 @@ def _get_or_build_posterior_predictive_summary(
         summary = posterior_predictive.get(cache_key)
         if summary is None and not include_draws:
             summary = posterior_predictive.get(draw_cache_key)
+        if summary is None:
+            summary = posterior_predictive.get(expt_name)
+            summary_x_axis = getattr(summary, 'x_axis_name', None)
+            if summary is not None and str(summary_x_axis) != str(x_axis_name):
+                summary = None
         if summary is not None:
             return summary
 
+        posterior_samples = getattr(fit_results, 'posterior_samples', None)
+        if posterior_samples is None:
+            return None
+
         summary = self._build_posterior_predictive_summary(
             fit_results=fit_results,
             experiment=experiment,
@@ -5537,7 +5775,7 @@ def _bragg_tick_d_spacing(
     def _plot_param_series_from_csv(
         self,
         csv_path: str,
-        unique_name: str,
+        column_names: str | list[str],
         param_descriptor: object,
         versus_path: str | None = None,
     ) -> None:
@@ -5545,10 +5783,10 @@ def _plot_param_series_from_csv(
         Plot a parameter's value across sequential fit results.
 
         Reads data from the CSV file at *csv_path*.  The y-axis values
-        come from the column named *unique_name*, uncertainties from
-        ``{unique_name}.uncertainty``. When *versus_path* is provided,
-        the x-axis uses the corresponding ``diffrn.*`` CSV column;
-        otherwise the row index is used.
+        come from the first matching column named in *column_names*,
+        with uncertainties from ``{column_name}.uncertainty``. When
+        *versus_path* is provided, the x-axis uses the corresponding
+        ``diffrn.*`` CSV column; otherwise the row index is used.
 
         Axis labels use the live parameter descriptor and, when
         available, a template diffrn descriptor resolved from
@@ -5558,8 +5796,8 @@ def _plot_param_series_from_csv(
         ----------
         csv_path : str
             Path to the ``results.csv`` file.
-        unique_name : str
-            Unique name of the parameter to plot (CSV column key).
+        column_names : str | list[str]
+            Candidate CSV column keys to plot.
         param_descriptor : object
             The live parameter descriptor (for axis label / units).
         versus_path : str | None, default=None
@@ -5568,16 +5806,23 @@ def _plot_param_series_from_csv(
         """
         df = pd.read_csv(csv_path)
 
-        if unique_name not in df.columns:
+        column_candidates = [column_names] if isinstance(column_names, str) else column_names
+
+        column_name = next((name for name in column_candidates if name in df.columns), None)
+        if column_name is None:
             log.warning(
-                f"Parameter '{unique_name}' not found in CSV columns. "
+                f"Parameter '{column_candidates[0]}' not found in CSV columns. "
                 f'Available: {list(df.columns)}'
             )
             return
 
-        y = df[unique_name].astype(float).tolist()
-        uncert_col = f'{unique_name}.uncertainty'
-        sy = df[uncert_col].astype(float).tolist() if uncert_col in df.columns else [0.0] * len(y)
+        y = self._numeric_series_values(df[column_name])
+        uncert_col = f'{column_name}.uncertainty'
+        sy = (
+            self._numeric_series_values(df[uncert_col])
+            if uncert_col in df.columns
+            else [0.0] * len(y)
+        )
 
         # X-axis: diffrn column or row index
         diffrn_col = versus_path
@@ -5594,7 +5839,7 @@ def _plot_param_series_from_csv(
         param_units = getattr(param_descriptor, 'units', '')
         y_label = f'Parameter value ({param_units})' if param_units else 'Parameter value'
 
-        title = f"Parameter '{unique_name}' across fit results"
+        title = f"Parameter '{column_name}' across fit results"
 
         self._backend.plot_scatter(
             x=x,
diff --git a/src/easydiffraction/display/progress.py b/src/easydiffraction/display/progress.py
index ef381c505..6123e626e 100644
--- a/src/easydiffraction/display/progress.py
+++ b/src/easydiffraction/display/progress.py
@@ -31,6 +31,8 @@
 
 ACTIVITY_LABEL_BURN_IN = 'Burn-in...'
 ACTIVITY_LABEL_FITTING = 'Fitting...'
+ACTIVITY_LABEL_POST_PROCESSING = 'Post-processing...'
+ACTIVITY_LABEL_PRE_PROCESSING = 'Pre-processing...'
 ACTIVITY_LABEL_PROCESSING = 'Processing...'
 ACTIVITY_LABEL_SAMPLING = 'Sampling...'
 ACTIVITY_ACCENT_COLOR = '#d97706'
diff --git a/src/easydiffraction/io/ascii.py b/src/easydiffraction/io/ascii.py
index faab76820..dcbea91b3 100644
--- a/src/easydiffraction/io/ascii.py
+++ b/src/easydiffraction/io/ascii.py
@@ -5,20 +5,22 @@
 from __future__ import annotations
 
 import re
-import tempfile
 import zipfile
 from io import StringIO
 from pathlib import Path
 
 import numpy as np
 
+from easydiffraction.utils.environment import create_artifact_temp_dir
+from easydiffraction.utils.environment import resolve_artifact_path
+
 
 def _resolve_extraction_destination(destination: str | Path | None) -> Path:
     """Return an extraction directory for ZIP contents."""
     if destination is None:
-        return Path(tempfile.mkdtemp(prefix='ed_zip_'))
+        return create_artifact_temp_dir(prefix='ed_zip_')
 
-    extract_dir = Path(destination)
+    extract_dir = resolve_artifact_path(destination)
     if not extract_dir.is_absolute():
         extract_dir = Path.cwd() / extract_dir
 
@@ -45,7 +47,9 @@ def extract_project_from_zip(
         Path to the ZIP archive containing the project.
     destination : str | Path | None, default=None
         Directory to extract into.  When ``None``, a temporary directory
-        is created.
+        is created. Relative destinations are resolved against the
+        configured artifact root when ``EASYDIFFRACTION_ARTIFACT_ROOT``
+        is set.
 
     Returns
     -------
@@ -65,11 +69,7 @@ def extract_project_from_zip(
         msg = f'ZIP file not found: {zip_path}'
         raise FileNotFoundError(msg)
 
-    if destination is not None:
-        extract_dir = Path(destination)
-        extract_dir.mkdir(parents=True, exist_ok=True)
-    else:
-        extract_dir = Path(tempfile.mkdtemp(prefix='ed_zip_'))
+    extract_dir = _resolve_extraction_destination(destination)
 
     with zipfile.ZipFile(zip_path, 'r') as zf:
         # Determine the project directory from the archive contents
@@ -107,7 +107,8 @@ def extract_data_paths_from_zip(
     destination : str | Path | None, default=None
         Directory to extract files into.  When ``None``, a temporary
         directory is created. Relative destinations are resolved against
-        the current working directory.
+        the current working directory, or against the configured
+        artifact root when ``EASYDIFFRACTION_ARTIFACT_ROOT`` is set.
 
     Returns
     -------
diff --git a/src/easydiffraction/io/cif/serialize.py b/src/easydiffraction/io/cif/serialize.py
index 5c3a29c14..ef9fa277d 100644
--- a/src/easydiffraction/io/cif/serialize.py
+++ b/src/easydiffraction/io/cif/serialize.py
@@ -3,6 +3,7 @@
 
 from __future__ import annotations
 
+import textwrap
 from typing import TYPE_CHECKING
 from typing import Any
 
@@ -21,15 +22,15 @@
     from easydiffraction.core.category import CategoryItem
     from easydiffraction.core.variable import GenericDescriptorBase
 
-# Maximum CIF description length before using semicolon-delimited block
-_CIF_DESCRIPTION_WRAP_LEN = 60
-
 # Minimum string length to check for surrounding quotes
 _MIN_QUOTED_LEN = 2
 
 # Number of significant digits kept for CIF uncertainty notation
 _CIF_UNCERTAINTY_SIG_DIGITS = 2
 
+# Maximum CIF description length before using semicolon-delimited block
+_CIF_DESCRIPTION_WRAP_LEN = 60
+
 
 def format_value(value: object) -> str:
     """
@@ -51,9 +52,9 @@ def format_value(value: object) -> str:
     # Booleans use CIF true/false tokens
     elif isinstance(value, bool):
         value = 'true' if value else 'false'
-    # Convert ints to floats
-    elif isinstance(value, int):
-        value = float(value)
+    # Preserve integers as integers in CIF output
+    elif isinstance(value, (int, np.integer)):
+        value = str(int(value))
     # Empty strings → CIF unknown marker
     elif isinstance(value, str) and not value.strip():
         value = '?'
@@ -79,6 +80,13 @@ def _strip_optional_quotes(raw: str) -> str:
     return raw[1:-1] if is_quoted else raw
 
 
+def _strip_cif_text_field_delimiters(raw: str) -> str:
+    """Return CIF text-field content without delimiter lines."""
+    if raw.startswith(';\n') and raw.endswith('\n;'):
+        return raw[2:-2].strip()
+    return raw
+
+
 def _parse_bool_cif_value(raw: str) -> bool | str:
     """Parse CIF boolean tokens, returning the raw token if invalid."""
     normalized_value = _strip_optional_quotes(raw).lower()
@@ -335,25 +343,38 @@ def datablock_collection_to_cif(collection: object) -> str:
     return '\n\n'.join([block.as_cif for block in collection.values()])
 
 
+def _format_project_description(description: str) -> str:
+    """Format project descriptions as CIF text."""
+    normalized_description = ' '.join(description.split())
+    if not normalized_description:
+        return '?'
+
+    if len(normalized_description) > _CIF_DESCRIPTION_WRAP_LEN:
+        wrapped_description = '\n'.join(
+            textwrap.wrap(
+                normalized_description,
+                width=_CIF_DESCRIPTION_WRAP_LEN,
+                break_long_words=False,
+                break_on_hyphens=False,
+            )
+        )
+        return f'\n;\n{wrapped_description}\n;'
+
+    return format_value(normalized_description)
+
+
 def project_info_to_cif(info: object) -> str:
     """Render ProjectInfo to CIF text (id, title, description)."""
     name = f'{info.name}'
 
     title = f'{info.title}'
     if ' ' in title:
-        title = f"'{title}'"
-
-    if len(info.description) > _CIF_DESCRIPTION_WRAP_LEN:
-        description = f'\n;\n{info.description}\n;'
-    elif info.description:
-        description = f'{info.description}'
-        if ' ' in description:
-            description = f"'{description}'"
-    else:
-        description = '?'
+        title = format_value(info.title)
+
+    description = _format_project_description(info.description)
 
-    created = f"'{info.created.strftime('%d %b %Y %H:%M:%S')}'"
-    last_modified = f"'{info.last_modified.strftime('%d %b %Y %H:%M:%S')}'"
+    created = format_value(info.created.strftime('%d %b %Y %H:%M:%S'))
+    last_modified = format_value(info.last_modified.strftime('%d %b %Y %H:%M:%S'))
 
     return (
         f'_project.id               {name}\n'
@@ -458,6 +479,13 @@ def _wrap_in_data_block(cif_text: str, block_name: str = '_') -> str:
     return f'data_{block_name}\n\n{cif_text}'
 
 
+def _project_block_from_cif_text(cif_text: str) -> gemmi.cif.Block:
+    """Parse project CIF text."""
+    import gemmi  # noqa: PLC0415
+
+    return gemmi.cif.read_string(_wrap_in_data_block(cif_text, 'project')).sole_block()
+
+
 def _populate_project_info_from_block(
     info: object,
     block: gemmi.cif.Block,
@@ -487,9 +515,7 @@ def project_info_from_cif(info: object, cif_text: str) -> None:
     """
     Populate a ProjectInfo instance from CIF text.
 
-    Reads ``_project.id``, ``_project.title``, and
-    ``_project.description`` from the given CIF string and sets them on
-    the *info* object.
+    Reads the core project metadata fields from CIF text.
 
     Parameters
     ----------
@@ -498,10 +524,7 @@ def project_info_from_cif(info: object, cif_text: str) -> None:
     cif_text : str
         CIF text content of ``project.cif``.
     """
-    import gemmi  # noqa: PLC0415
-
-    doc = gemmi.cif.read_string(_wrap_in_data_block(cif_text, 'project'))
-    block = doc.sole_block()
+    block = _project_block_from_cif_text(cif_text)
 
     _populate_project_info_from_block(info, block)
 
@@ -510,10 +533,7 @@ def project_config_from_cif(project: object, cif_text: str) -> None:
     """
     Populate project-level configuration from ``project.cif`` text.
     """
-    import gemmi  # noqa: PLC0415
-
-    doc = gemmi.cif.read_string(_wrap_in_data_block(cif_text, 'project'))
-    block = doc.sole_block()
+    block = _project_block_from_cif_text(cif_text)
 
     _populate_project_info_from_block(project.info, block)
 
@@ -521,6 +541,10 @@ def project_config_from_cif(project: object, cif_text: str) -> None:
     if rendering is not None:
         rendering.from_cif(block)
 
+    verbosity = getattr(project, 'verbosity', None)
+    if verbosity is not None:
+        verbosity.from_cif(block)
+
 
 def analysis_from_cif(analysis: object, cif_text: str) -> None:
     """
@@ -556,6 +580,82 @@ def analysis_from_cif(analysis: object, cif_text: str) -> None:
     if analysis.constraints._items:
         analysis.constraints.enable()
 
+    if _has_persisted_fit_state_sections(block):
+        _restore_persisted_fit_state(analysis, block)
+
+
+def _has_persisted_fit_state_sections(block: object) -> bool:
+    """Return True when any persisted fit-state section is present."""
+    scalar_tags = (
+        '_fit_result.result_kind',
+        '_deterministic_result.optimizer_name',
+        '_bayesian_result.sampler_name',
+        '_bayesian_sampler.steps',
+        '_bayesian_convergence.converged',
+    )
+    loop_tags = (
+        '_fit_parameter.param_unique_name',
+        '_fit_parameter_correlation.param_unique_name_i',
+        '_bayesian_parameter_posterior.unique_name',
+        '_bayesian_distribution_cache.param_unique_name',
+        '_bayesian_pair_cache.param_unique_name_x',
+        '_bayesian_predictive_dataset.experiment_name',
+    )
+
+    return any(_has_cif_value(block, tag) for tag in scalar_tags) or any(
+        _has_cif_loop(block, tag) for tag in loop_tags
+    )
+
+
+def _restore_common_fit_state(analysis: object, block: object) -> None:
+    """Restore fit-state categories shared by both fit kinds."""
+    analysis.fit_parameters.from_cif(block)
+    analysis.fit_result.from_cif(block)
+    analysis.fit_parameter_correlations.from_cif(block)
+
+
+def _restore_deterministic_fit_state(analysis: object, block: object) -> None:
+    """Restore deterministic-only persisted fit-state categories."""
+    analysis.deterministic_result.from_cif(block)
+
+
+def _restore_bayesian_fit_state(analysis: object, block: object) -> None:
+    """Restore Bayesian-only persisted fit-state categories."""
+    analysis.bayesian_result.from_cif(block)
+    analysis.bayesian_sampler.from_cif(block)
+    analysis.bayesian_convergence.from_cif(block)
+    analysis.bayesian_parameter_posteriors.from_cif(block)
+    analysis.bayesian_distribution_caches.from_cif(block)
+    analysis.bayesian_pair_caches.from_cif(block)
+    analysis.bayesian_predictive_datasets.from_cif(block)
+    analysis._sync_live_minimizer_from_persisted_fit_state()
+
+
+def _restore_persisted_fit_state(analysis: object, block: object) -> None:
+    """
+    Restore persisted fit-state categories after analysis configuration.
+    """
+    from easydiffraction.analysis.enums import FitResultKindEnum  # noqa: PLC0415
+
+    analysis._set_has_persisted_fit_state(value=True)
+    _restore_common_fit_state(analysis, block)
+
+    result_kind_value = analysis.fit_result.result_kind.value
+    try:
+        result_kind = FitResultKindEnum(result_kind_value)
+    except ValueError:
+        log.warning(
+            'Unsupported _fit_result.result_kind in analysis CIF: '
+            f'{result_kind_value!r}. Skipping kind-specific fit-state categories.',
+        )
+        return
+
+    if result_kind is FitResultKindEnum.DETERMINISTIC:
+        _restore_deterministic_fit_state(analysis, block)
+        return
+
+    _restore_bayesian_fit_state(analysis, block)
+
 
 def _collect_legacy_analysis_tags(block: object) -> list[str]:
     """Return deprecated analysis CIF tags present in a block."""
@@ -687,10 +787,8 @@ def _read(tag: str) -> str | None:
         # CIF unknown / inapplicable markers
         if raw in {'?', '.'}:
             return None
-        # Strip surrounding quotes
-        if len(raw) >= _MIN_QUOTED_LEN and raw[0] == raw[-1] and raw[0] in {"'", '"'}:
-            raw = raw[1:-1]
-        return raw
+        raw = _strip_cif_text_field_delimiters(raw)
+        return _strip_optional_quotes(raw)
 
     return _read
 
@@ -747,33 +845,9 @@ def param_from_cif(
     if not found_values:
         return
 
-    # If found, pick the one at the given index
+    # If found, pick the one at the given index.
     raw = found_values[idx]
-
-    # CIF unknown / inapplicable markers → keep default
-    if raw in {'?', '.'}:
-        return
-
-    # If numeric, parse with uncertainty if present
-    if self._value_type == DataTypes.NUMERIC:
-        has_brackets = '(' in raw
-        u = str_to_ufloat(raw)
-        self.value = u.n
-        if has_brackets and hasattr(self, 'free'):
-            self.free = True  # type: ignore[attr-defined]
-            if not np.isnan(u.s) and hasattr(self, 'uncertainty'):
-                self.uncertainty = u.s  # type: ignore[attr-defined]
-
-    # If string, strip quotes if present
-    elif self._value_type == DataTypes.STRING:
-        self.value = _strip_optional_quotes(raw)
-
-    elif self._value_type == DataTypes.BOOL:
-        self.value = _parse_bool_cif_value(raw)
-
-    # Other types are not supported
-    else:
-        log.debug(f'Unrecognized type: {self._value_type}')
+    _set_param_from_raw_cif_value(self, raw)
 
 
 def category_item_from_cif(
@@ -803,11 +877,23 @@ def _set_param_from_raw_cif_value(
     raw : str
         The raw string from the CIF loop cell.
     """
+    raw = _strip_cif_text_field_delimiters(raw)
+
     # CIF unknown / inapplicable markers → keep default
     if raw in {'?', '.'}:
         return
 
-    if param._value_type == DataTypes.NUMERIC:
+    if param._value_type == DataTypes.INTEGER:
+        numeric_value = str_to_ufloat(raw).n
+        integer_value = round(numeric_value)
+        if not np.isclose(numeric_value, integer_value):
+            log.warning(
+                f'Ignoring non-integer CIF value {raw!r} for integer field {param.unique_name}.'
+            )
+            return
+        param.value = integer_value
+
+    elif param._value_type == DataTypes.NUMERIC:
         has_brackets = '(' in raw
         u = str_to_ufloat(raw)
         param.value = u.n
diff --git a/src/easydiffraction/io/results_sidecar.py b/src/easydiffraction/io/results_sidecar.py
new file mode 100644
index 000000000..36d442c77
--- /dev/null
+++ b/src/easydiffraction/io/results_sidecar.py
@@ -0,0 +1,652 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Bayesian fit sidecar read/write helpers."""
+
+from __future__ import annotations
+
+from pathlib import Path
+from tempfile import NamedTemporaryFile
+
+import numpy as np
+
+from easydiffraction.utils.logging import log
+
+_DEFAULT_SIDECAR_FILE_NAME = 'results.h5'
+_POSTERIOR_PARAMETER_SAMPLES_PATH = '/posterior/parameter_samples'
+_POSTERIOR_LOG_POSTERIOR_PATH = '/posterior/log_posterior'
+_POSTERIOR_DRAW_INDEX_PATH = '/posterior/draw_index'
+_POSTERIOR_SAMPLE_NDIM = 3
+_PREDICTIVE_DRAWS_NDIM = 2
+
+
+def _normalized_hdf5_path(path: str) -> str:
+    """Return an HDF5 path without a leading slash."""
+    return path.lstrip('/')
+
+
+def _sidecar_file_name(analysis: object) -> str:
+    """Return the configured sidecar file name for an analysis."""
+    bayesian_result = getattr(analysis, 'bayesian_result', None)
+    if bayesian_result is None:
+        return _DEFAULT_SIDECAR_FILE_NAME
+
+    file_name = bayesian_result.sidecar_file.value
+    if not isinstance(file_name, str) or not file_name.strip():
+        return _DEFAULT_SIDECAR_FILE_NAME
+
+    normalized_name = file_name.strip()
+    normalized_path = Path(normalized_name)
+    if (
+        normalized_path.is_absolute()
+        or normalized_path.name in {'', '.', '..'}
+        or normalized_path.name != normalized_name
+    ):
+        log.warning(
+            'Ignoring Bayesian sidecar file path outside the analysis directory: '
+            f'{normalized_name!r}. Using {_DEFAULT_SIDECAR_FILE_NAME!r} instead.'
+        )
+        return _DEFAULT_SIDECAR_FILE_NAME
+
+    return normalized_path.name
+
+
+def _sidecar_path(*, analysis: object, analysis_dir: Path) -> Path:
+    """Return the results sidecar path inside the analysis directory."""
+    resolved_analysis_dir = analysis_dir.resolve()
+    sidecar_path = (resolved_analysis_dir / _sidecar_file_name(analysis)).resolve()
+    if sidecar_path.parent != resolved_analysis_dir:
+        log.warning(
+            'Resolved Bayesian sidecar file path escaped the analysis directory. '
+            f'Using {_DEFAULT_SIDECAR_FILE_NAME!r} instead.'
+        )
+        return resolved_analysis_dir / _DEFAULT_SIDECAR_FILE_NAME
+    return sidecar_path
+
+
+def _should_use_sidecar(analysis: object) -> bool:
+    """
+    Return whether the analysis currently expects a Bayesian sidecar.
+    """
+    has_fit_state = getattr(analysis, '_has_persisted_fit_state', None)
+    if not callable(has_fit_state) or not has_fit_state():
+        return False
+
+    if analysis.fit_result.result_kind.value != 'bayesian':
+        return False
+
+    return any((
+        analysis.bayesian_result.has_posterior_samples.value,
+        len(analysis.bayesian_distribution_caches) > 0,
+        len(analysis.bayesian_pair_caches) > 0,
+        len(analysis.bayesian_predictive_datasets) > 0,
+    ))
+
+
+def _delete_stale_sidecar(sidecar_path: Path) -> None:
+    """
+    Delete an existing sidecar when no persisted arrays should remain.
+    """
+    if sidecar_path.is_file():
+        sidecar_path.unlink()
+
+
+def _create_dataset(handle: object, path: str, data: np.ndarray) -> None:
+    """Create or replace one dataset in an open HDF5 file."""
+    normalized_path = _normalized_hdf5_path(path)
+    group_name, _, dataset_name = normalized_path.rpartition('/')
+    group = handle.require_group(group_name) if group_name else handle
+    if dataset_name in group:
+        del group[dataset_name]
+    group.create_dataset(dataset_name, data=data)
+
+
+def _read_dataset(handle: object, path: str) -> np.ndarray | None:
+    """Read one dataset from an open HDF5 file when it exists."""
+    normalized_path = _normalized_hdf5_path(path)
+    if normalized_path not in handle:
+        return None
+    return np.asarray(handle[normalized_path])
+
+
+def _posterior_payload_from_analysis(analysis: object) -> dict[str, np.ndarray | None]:
+    """Return posterior arrays from runtime or restored data."""
+    fit_results = getattr(analysis, 'fit_results', None)
+    posterior_samples = getattr(fit_results, 'posterior_samples', None)
+    if posterior_samples is not None:
+        return {
+            'parameter_samples': np.asarray(posterior_samples.parameter_samples, dtype=float),
+            'log_posterior': (
+                None
+                if posterior_samples.log_posterior is None
+                else np.asarray(posterior_samples.log_posterior, dtype=float)
+            ),
+            'draw_index': (
+                None
+                if posterior_samples.draw_index is None
+                else np.asarray(posterior_samples.draw_index)
+            ),
+        }
+
+    sidecar_data = getattr(analysis, '_persisted_fit_state_sidecar', {})
+    return dict(sidecar_data.get('posterior', {}))
+
+
+def _distribution_cache_payload(analysis: object) -> dict[str, dict[str, np.ndarray]]:
+    """Return persisted distribution caches keyed by parameter name."""
+    sidecar_data = getattr(analysis, '_persisted_fit_state_sidecar', {})
+    return dict(sidecar_data.get('distribution_caches', {}))
+
+
+def _pair_cache_payload(analysis: object) -> dict[str, dict[str, np.ndarray]]:
+    """Return persisted pair-cache arrays keyed by cache id."""
+    sidecar_data = getattr(analysis, '_persisted_fit_state_sidecar', {})
+    return dict(sidecar_data.get('pair_caches', {}))
+
+
+def _predictive_payload(analysis: object) -> dict[str, dict[str, np.ndarray]]:
+    """Return persisted predictive arrays keyed by experiment name."""
+    fit_results = getattr(analysis, 'fit_results', None)
+    posterior_predictive = getattr(fit_results, 'posterior_predictive', None)
+    if posterior_predictive:
+        payload: dict[str, dict[str, np.ndarray]] = {}
+        for runtime_key, summary in posterior_predictive.items():
+            experiment_name = getattr(summary, 'experiment_name', None)
+            if not isinstance(experiment_name, str) or not experiment_name.strip():
+                experiment_name = runtime_key
+
+            dataset_payload = payload.setdefault(experiment_name, {})
+            dataset_payload['x'] = np.asarray(summary.x, dtype=float)
+            dataset_payload['best_sample_prediction'] = np.asarray(
+                summary.best_sample_prediction,
+                dtype=float,
+            )
+            if summary.lower_95 is not None:
+                dataset_payload['lower_95'] = np.asarray(summary.lower_95, dtype=float)
+            if summary.upper_95 is not None:
+                dataset_payload['upper_95'] = np.asarray(summary.upper_95, dtype=float)
+            if summary.lower_68 is not None:
+                dataset_payload['lower_68'] = np.asarray(summary.lower_68, dtype=float)
+            if summary.upper_68 is not None:
+                dataset_payload['upper_68'] = np.asarray(summary.upper_68, dtype=float)
+            if summary.draws is not None:
+                dataset_payload['draws'] = np.asarray(summary.draws, dtype=float)
+        return payload
+
+    sidecar_data = getattr(analysis, '_persisted_fit_state_sidecar', {})
+    return dict(sidecar_data.get('predictive_datasets', {}))
+
+
+def _validate_posterior_payload(
+    analysis: object,
+    payload: dict[str, np.ndarray | None],
+) -> bool:
+    """Return whether posterior arrays match stored metadata."""
+    parameter_samples = payload.get('parameter_samples')
+    if parameter_samples is None:
+        if analysis.bayesian_result.has_posterior_samples.value:
+            log.warning('Bayesian fit-state expects posterior samples, but none are available.')
+        return False
+
+    parameter_samples = np.asarray(parameter_samples, dtype=float)
+    if parameter_samples.ndim != _POSTERIOR_SAMPLE_NDIM:
+        log.warning(
+            'Posterior parameter samples must have shape (n_draws, n_chains, n_parameters).'
+        )
+        return False
+
+    n_draws, n_chains, n_parameters = parameter_samples.shape
+    if not _posterior_manifest_counts_match(
+        analysis,
+        n_draws=n_draws,
+        n_chains=n_chains,
+        n_parameters=n_parameters,
+    ):
+        return False
+
+    return _posterior_aux_shapes_match(
+        payload,
+        n_draws=n_draws,
+        n_chains=n_chains,
+    )
+
+
+def _posterior_manifest_counts_match(
+    analysis: object,
+    *,
+    n_draws: int,
+    n_chains: int,
+    n_parameters: int,
+) -> bool:
+    """Return whether manifest counts match the sample shape."""
+    if analysis.bayesian_convergence.n_draws.value not in {0, n_draws}:
+        log.warning('Posterior sample draw count does not match bayesian_convergence.n_draws.')
+        return False
+    if analysis.bayesian_convergence.n_chains.value not in {0, n_chains}:
+        log.warning('Posterior sample chain count does not match bayesian_convergence.n_chains.')
+        return False
+    if analysis.bayesian_convergence.n_parameters.value not in {0, n_parameters}:
+        log.warning(
+            'Posterior sample parameter count does not match bayesian_convergence.n_parameters.'
+        )
+        return False
+
+    return True
+
+
+def _posterior_aux_shapes_match(
+    payload: dict[str, np.ndarray | None],
+    *,
+    n_draws: int,
+    n_chains: int,
+) -> bool:
+    """Return whether auxiliary arrays match the sample shape."""
+    log_posterior = payload.get('log_posterior')
+    if log_posterior is not None and np.asarray(log_posterior).shape != (n_draws, n_chains):
+        log.warning(
+            'Posterior log-posterior array does not match posterior sample draw and chain axes.'
+        )
+        return False
+
+    draw_index = payload.get('draw_index')
+    if draw_index is not None and np.asarray(draw_index).shape != (n_draws,):
+        log.warning('Posterior draw-index array does not match posterior sample draw count.')
+        return False
+
+    return True
+
+
+def _write_posterior_payload(handle: object, analysis: object) -> bool:
+    """Write canonical posterior arrays when they are available."""
+    payload = _posterior_payload_from_analysis(analysis)
+    if not _validate_posterior_payload(analysis, payload):
+        return False
+
+    parameter_samples = np.asarray(payload['parameter_samples'], dtype=float)
+    _create_dataset(handle, _POSTERIOR_PARAMETER_SAMPLES_PATH, parameter_samples)
+
+    log_posterior = payload.get('log_posterior')
+    if log_posterior is not None:
+        _create_dataset(handle, _POSTERIOR_LOG_POSTERIOR_PATH, np.asarray(log_posterior))
+
+    draw_index = payload.get('draw_index')
+    if draw_index is not None:
+        _create_dataset(handle, _POSTERIOR_DRAW_INDEX_PATH, np.asarray(draw_index))
+
+    return True
+
+
+def _write_distribution_caches(handle: object, analysis: object) -> bool:
+    """Write cached posterior distribution arrays for manifest rows."""
+    payload = _distribution_cache_payload(analysis)
+    wrote_any = False
+    for cache in analysis.bayesian_distribution_caches:
+        cache_data = payload.get(cache.param_unique_name.value)
+        if cache_data is None:
+            continue
+
+        x_values = np.asarray(cache_data.get('x'))
+        density_values = np.asarray(cache_data.get('density'))
+        n_grid = int(cache.n_grid.value)
+        if x_values.shape != (n_grid,) or density_values.shape != (n_grid,):
+            log.warning(
+                'Skipping Bayesian distribution cache with shape mismatch for '
+                f'{cache.param_unique_name.value!r}.'
+            )
+            continue
+
+        _create_dataset(handle, cache.x_path.value, x_values)
+        _create_dataset(handle, cache.density_path.value, density_values)
+        wrote_any = True
+
+    return wrote_any
+
+
+def _write_pair_caches(handle: object, analysis: object) -> bool:
+    """Write cached posterior pair-density arrays for manifest rows."""
+    payload = _pair_cache_payload(analysis)
+    wrote_any = False
+    for cache in analysis.bayesian_pair_caches:
+        cache_data = payload.get(cache.id.value)
+        if cache_data is None:
+            continue
+
+        x_values = np.asarray(cache_data.get('x'))
+        y_values = np.asarray(cache_data.get('y'))
+        density_values = np.asarray(cache_data.get('density'))
+        contour_levels = np.asarray(cache_data.get('contour_levels'))
+        n_grid_x = int(cache.n_grid_x.value)
+        n_grid_y = int(cache.n_grid_y.value)
+
+        valid_density_shape = density_values.shape in {
+            (n_grid_y, n_grid_x),
+            (n_grid_x, n_grid_y),
+        }
+        if (
+            x_values.shape != (n_grid_x,)
+            or y_values.shape != (n_grid_y,)
+            or not valid_density_shape
+        ):
+            log.warning(
+                f'Skipping Bayesian pair cache with shape mismatch for {cache.id.value!r}.'
+            )
+            continue
+
+        _create_dataset(handle, cache.x_path.value, x_values)
+        _create_dataset(handle, cache.y_path.value, y_values)
+        _create_dataset(handle, cache.density_path.value, density_values)
+        _create_dataset(handle, cache.contour_level_path.value, contour_levels)
+        wrote_any = True
+
+    return wrote_any
+
+
+def _write_predictive_datasets(handle: object, analysis: object) -> bool:
+    """Write cached posterior predictive arrays for manifest rows."""
+    payload = _predictive_payload(analysis)
+    wrote_any = False
+    for dataset in analysis.bayesian_predictive_datasets:
+        dataset_data = payload.get(dataset.experiment_name.value)
+        if dataset_data is None:
+            continue
+
+        x_values = np.asarray(dataset_data.get('x'))
+        best_sample_prediction = np.asarray(dataset_data.get('best_sample_prediction'))
+        n_x = int(dataset.n_x.value)
+        if x_values.shape != (n_x,) or best_sample_prediction.shape != (n_x,):
+            log.warning(
+                'Skipping Bayesian predictive dataset with shape mismatch for '
+                f'{dataset.experiment_name.value!r}.'
+            )
+            continue
+
+        _create_dataset(handle, dataset.x_path.value, x_values)
+        _create_dataset(
+            handle,
+            dataset.best_sample_prediction_path.value,
+            best_sample_prediction,
+        )
+
+        for field_name, path_value in (
+            ('lower_95', dataset.lower_95_path.value),
+            ('upper_95', dataset.upper_95_path.value),
+            ('lower_68', dataset.lower_68_path.value),
+            ('upper_68', dataset.upper_68_path.value),
+        ):
+            values = dataset_data.get(field_name)
+            if values is None or path_value is None:
+                continue
+            values_array = np.asarray(values)
+            if values_array.shape != (n_x,):
+                log.warning(
+                    'Skipping Bayesian predictive band with shape mismatch for '
+                    f'{dataset.experiment_name.value!r}:{field_name}.'
+                )
+                continue
+            _create_dataset(handle, path_value, values_array)
+
+        draws = dataset_data.get('draws')
+        if draws is not None and dataset.draws_path.value is not None:
+            draws_array = np.asarray(draws)
+            if draws_array.ndim != _PREDICTIVE_DRAWS_NDIM or draws_array.shape[1] != n_x:
+                log.warning(
+                    'Skipping Bayesian predictive draws with shape mismatch for '
+                    f'{dataset.experiment_name.value!r}.'
+                )
+            elif dataset.n_draws_cached.value not in {0, draws_array.shape[0]}:
+                log.warning(
+                    'Skipping Bayesian predictive draws whose draw count does not match '
+                    'the manifest metadata.'
+                )
+            else:
+                _create_dataset(handle, dataset.draws_path.value, draws_array)
+
+        wrote_any = True
+
+    return wrote_any
+
+
+def write_analysis_results_sidecar(
+    *,
+    analysis: object,
+    analysis_dir: Path,
+) -> None:
+    """
+    Write persisted Bayesian arrays to ``analysis/results.h5``.
+
+    Parameters
+    ----------
+    analysis : object
+        Analysis instance that owns fit-state categories and runtime fit
+        results.
+    analysis_dir : Path
+        The project ``analysis/`` directory.
+
+    Raises
+    ------
+    Exception
+        Propagated when sidecar writing fails after temporary-file
+        cleanup.
+    """
+    sidecar_path = _sidecar_path(analysis=analysis, analysis_dir=analysis_dir)
+    if not _should_use_sidecar(analysis):
+        _delete_stale_sidecar(sidecar_path)
+        return
+
+    import h5py  # noqa: PLC0415
+
+    analysis_dir.mkdir(parents=True, exist_ok=True)
+    with NamedTemporaryFile(
+        delete=False,
+        dir=analysis_dir,
+        prefix=f'{sidecar_path.stem}.',
+        suffix=sidecar_path.suffix,
+    ) as temporary_file:
+        temporary_path = Path(temporary_file.name)
+
+    try:
+        with h5py.File(temporary_path, 'w') as handle:
+            wrote_any = _write_posterior_payload(handle, analysis)
+            wrote_any = _write_distribution_caches(handle, analysis) or wrote_any
+            wrote_any = _write_pair_caches(handle, analysis) or wrote_any
+            wrote_any = _write_predictive_datasets(handle, analysis) or wrote_any
+    except Exception:
+        if temporary_path.exists():
+            temporary_path.unlink()
+        raise
+
+    if not wrote_any:
+        temporary_path.unlink()
+        _delete_stale_sidecar(sidecar_path)
+        return
+
+    temporary_path.replace(sidecar_path)
+
+
+def _read_posterior_payload(handle: object, analysis: object) -> dict[str, np.ndarray]:
+    """Read canonical posterior arrays from a sidecar file."""
+    parameter_samples = _read_dataset(handle, _POSTERIOR_PARAMETER_SAMPLES_PATH)
+    if parameter_samples is None:
+        return {}
+
+    payload: dict[str, np.ndarray] = {
+        'parameter_samples': np.asarray(parameter_samples, dtype=float),
+    }
+    log_posterior = _read_dataset(handle, _POSTERIOR_LOG_POSTERIOR_PATH)
+    if log_posterior is not None:
+        payload['log_posterior'] = np.asarray(log_posterior, dtype=float)
+    draw_index = _read_dataset(handle, _POSTERIOR_DRAW_INDEX_PATH)
+    if draw_index is not None:
+        payload['draw_index'] = np.asarray(draw_index)
+
+    if not _validate_posterior_payload(analysis, payload):
+        return {}
+    return payload
+
+
+def _read_distribution_caches(
+    handle: object, analysis: object
+) -> dict[str, dict[str, np.ndarray]]:
+    """Read cached posterior distribution arrays for manifest rows."""
+    payload: dict[str, dict[str, np.ndarray]] = {}
+    for cache in analysis.bayesian_distribution_caches:
+        x_values = _read_dataset(handle, cache.x_path.value)
+        density_values = _read_dataset(handle, cache.density_path.value)
+        if x_values is None or density_values is None:
+            continue
+        if x_values.shape != (int(cache.n_grid.value),) or density_values.shape != (
+            int(cache.n_grid.value),
+        ):
+            log.warning(
+                'Skipping restored Bayesian distribution cache with shape mismatch for '
+                f'{cache.param_unique_name.value!r}.'
+            )
+            continue
+        payload[cache.param_unique_name.value] = {
+            'x': np.asarray(x_values),
+            'density': np.asarray(density_values),
+        }
+    return payload
+
+
+def _read_pair_caches(handle: object, analysis: object) -> dict[str, dict[str, np.ndarray]]:
+    """Read cached posterior pair-density arrays for manifest rows."""
+    payload: dict[str, dict[str, np.ndarray]] = {}
+    for cache in analysis.bayesian_pair_caches:
+        x_values = _read_dataset(handle, cache.x_path.value)
+        y_values = _read_dataset(handle, cache.y_path.value)
+        density_values = _read_dataset(handle, cache.density_path.value)
+        contour_levels = _read_dataset(handle, cache.contour_level_path.value)
+        if any(value is None for value in (x_values, y_values, density_values, contour_levels)):
+            continue
+
+        n_grid_x = int(cache.n_grid_x.value)
+        n_grid_y = int(cache.n_grid_y.value)
+        valid_density_shape = density_values.shape in {
+            (n_grid_y, n_grid_x),
+            (n_grid_x, n_grid_y),
+        }
+        if (
+            x_values.shape != (n_grid_x,)
+            or y_values.shape != (n_grid_y,)
+            or not valid_density_shape
+        ):
+            log.warning(
+                'Skipping restored Bayesian pair cache with shape mismatch for '
+                f'{cache.id.value!r}.'
+            )
+            continue
+
+        payload[cache.id.value] = {
+            'x': np.asarray(x_values),
+            'y': np.asarray(y_values),
+            'density': np.asarray(density_values),
+            'contour_levels': np.asarray(contour_levels),
+        }
+    return payload
+
+
+def _read_predictive_datasets(
+    handle: object, analysis: object
+) -> dict[str, dict[str, np.ndarray]]:
+    """Read cached posterior predictive arrays for manifest rows."""
+    payload: dict[str, dict[str, np.ndarray]] = {}
+    for dataset in analysis.bayesian_predictive_datasets:
+        x_values = _read_dataset(handle, dataset.x_path.value)
+        best_sample_prediction = _read_dataset(handle, dataset.best_sample_prediction_path.value)
+        if x_values is None or best_sample_prediction is None:
+            continue
+
+        n_x = int(dataset.n_x.value)
+        if x_values.shape != (n_x,) or best_sample_prediction.shape != (n_x,):
+            log.warning(
+                'Skipping restored Bayesian predictive dataset with shape mismatch for '
+                f'{dataset.experiment_name.value!r}.'
+            )
+            continue
+
+        dataset_payload: dict[str, np.ndarray] = {
+            'x': np.asarray(x_values),
+            'best_sample_prediction': np.asarray(best_sample_prediction),
+        }
+
+        for field_name, path_value in (
+            ('lower_95', dataset.lower_95_path.value),
+            ('upper_95', dataset.upper_95_path.value),
+            ('lower_68', dataset.lower_68_path.value),
+            ('upper_68', dataset.upper_68_path.value),
+            ('draws', dataset.draws_path.value),
+        ):
+            if path_value is None:
+                continue
+            values = _read_dataset(handle, path_value)
+            if values is None:
+                continue
+            values_array = np.asarray(values)
+            if field_name == 'draws':
+                if values_array.ndim != _PREDICTIVE_DRAWS_NDIM or values_array.shape[1] != n_x:
+                    log.warning(
+                        'Skipping restored Bayesian predictive draws with shape mismatch for '
+                        f'{dataset.experiment_name.value!r}.'
+                    )
+                    continue
+            elif values_array.shape != (n_x,):
+                log.warning(
+                    'Skipping restored Bayesian predictive band with shape mismatch for '
+                    f'{dataset.experiment_name.value!r}:{field_name}.'
+                )
+                continue
+            dataset_payload[field_name] = values_array
+
+        payload[dataset.experiment_name.value] = dataset_payload
+    return payload
+
+
+def read_analysis_results_sidecar(
+    *,
+    analysis: object,
+    analysis_dir: Path,
+) -> None:
+    """
+    Read persisted Bayesian arrays from ``analysis/results.h5``.
+
+    Parameters
+    ----------
+    analysis : object
+        Analysis instance that owns fit-state categories.
+    analysis_dir : Path
+        The project ``analysis/`` directory.
+    """
+    analysis._persisted_fit_state_sidecar = {}
+    if not _should_use_sidecar(analysis):
+        return
+
+    sidecar_path = _sidecar_path(analysis=analysis, analysis_dir=analysis_dir)
+    if not sidecar_path.is_file():
+        log.warning(
+            'Expected Bayesian results sidecar is missing: '
+            f"'{sidecar_path}'. Restoring available CIF summaries only."
+        )
+        return
+
+    import h5py  # noqa: PLC0415
+
+    with h5py.File(sidecar_path, 'r') as handle:
+        sidecar_data: dict[str, object] = {}
+
+        posterior_payload = _read_posterior_payload(handle, analysis)
+        if posterior_payload:
+            sidecar_data['posterior'] = posterior_payload
+
+        distribution_caches = _read_distribution_caches(handle, analysis)
+        if distribution_caches:
+            sidecar_data['distribution_caches'] = distribution_caches
+
+        pair_caches = _read_pair_caches(handle, analysis)
+        if pair_caches:
+            sidecar_data['pair_caches'] = pair_caches
+
+        predictive_datasets = _read_predictive_datasets(handle, analysis)
+        if predictive_datasets:
+            sidecar_data['predictive_datasets'] = predictive_datasets
+
+    analysis._persisted_fit_state_sidecar = sidecar_data
diff --git a/src/easydiffraction/project/categories/rendering/default.py b/src/easydiffraction/project/categories/rendering/default.py
index d3e2fad3e..179162a08 100644
--- a/src/easydiffraction/project/categories/rendering/default.py
+++ b/src/easydiffraction/project/categories/rendering/default.py
@@ -19,6 +19,10 @@
 from easydiffraction.utils.logging import console
 from easydiffraction.utils.utils import render_table
 
+AUTO_ENGINE = 'auto'
+CHART_ENGINE_OPTIONS = [AUTO_ENGINE, *[member.value for member in PlotterEngineEnum]]
+TABLE_ENGINE_OPTIONS = [AUTO_ENGINE, *[member.value for member in TableEngineEnum]]
+
 
 @RenderingFactory.register
 class Rendering(CategoryItem):
@@ -37,13 +41,14 @@ def __init__(self) -> None:
         self._plotter = Plotter()
         self._tabler = TableRenderer.get()
 
+        # Persist symbolic "auto" so project.cif stays portable.
         self._chart_engine = StringDescriptor(
             name='chart_engine',
             description='Chart renderer backend type',
             value_spec=AttributeSpec(
-                default=self._plotter.engine,
+                default=AUTO_ENGINE,
                 validator=MembershipValidator(
-                    allowed=[member.value for member in PlotterEngineEnum],
+                    allowed=CHART_ENGINE_OPTIONS,
                 ),
             ),
             cif_handler=CifHandler(names=['_rendering.chart_engine']),
@@ -52,14 +57,46 @@ def __init__(self) -> None:
             name='table_engine',
             description='Table renderer backend type',
             value_spec=AttributeSpec(
-                default=self._tabler.engine,
+                default=AUTO_ENGINE,
                 validator=MembershipValidator(
-                    allowed=[member.value for member in TableEngineEnum],
+                    allowed=TABLE_ENGINE_OPTIONS,
                 ),
             ),
             cif_handler=CifHandler(names=['_rendering.table_engine']),
         )
 
+    @staticmethod
+    def _resolved_chart_engine(value: str) -> str:
+        if value == AUTO_ENGINE:
+            return PlotterEngineEnum.default().value
+        return value
+
+    @staticmethod
+    def _resolved_table_engine(value: str) -> str:
+        if value == AUTO_ENGINE:
+            return TableEngineEnum.default().value
+        return value
+
+    def _set_chart_engine(self, value: str) -> None:
+        if value not in CHART_ENGINE_OPTIONS:
+            self._plotter.engine = value
+            return
+
+        resolved_engine = self._resolved_chart_engine(value)
+        if self._plotter.engine != resolved_engine:
+            self._plotter.engine = resolved_engine
+        self._chart_engine.value = value
+
+    def _set_table_engine(self, value: str) -> None:
+        if value not in TABLE_ENGINE_OPTIONS:
+            self._tabler.engine = value
+            return
+
+        resolved_engine = self._resolved_table_engine(value)
+        if self._tabler.engine != resolved_engine:
+            self._tabler.engine = resolved_engine
+        self._table_engine.value = value
+
     @property
     def chart_engine(self) -> StringDescriptor:
         """Chart renderer backend type."""
@@ -67,8 +104,7 @@ def chart_engine(self) -> StringDescriptor:
 
     @chart_engine.setter
     def chart_engine(self, value: str) -> None:
-        self._plotter.engine = value
-        self._chart_engine.value = self._plotter.engine
+        self._set_chart_engine(value)
 
     @property
     def table_engine(self) -> StringDescriptor:
@@ -77,8 +113,7 @@ def table_engine(self) -> StringDescriptor:
 
     @table_engine.setter
     def table_engine(self, value: str) -> None:
-        self._tabler.engine = value
-        self._table_engine.value = self._tabler.engine
+        self._set_table_engine(value)
 
     @property
     def plotter(self) -> Plotter:
@@ -123,14 +158,8 @@ def from_cif(self, block: object, idx: int = 0) -> None:
         del idx
         chart_engine = read_cif_str(block, '_rendering.chart_engine')
         if chart_engine is not None:
-            if chart_engine == self._plotter.engine:
-                self._chart_engine.value = chart_engine
-            else:
-                self.chart_engine = chart_engine
+            self._set_chart_engine(chart_engine)
 
         table_engine = read_cif_str(block, '_rendering.table_engine')
         if table_engine is not None:
-            if table_engine == self._tabler.engine:
-                self._table_engine.value = table_engine
-            else:
-                self.table_engine = table_engine
+            self._set_table_engine(table_engine)
diff --git a/src/easydiffraction/project/categories/verbosity/__init__.py b/src/easydiffraction/project/categories/verbosity/__init__.py
new file mode 100644
index 000000000..a7bfce5b3
--- /dev/null
+++ b/src/easydiffraction/project/categories/verbosity/__init__.py
@@ -0,0 +1,8 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Project verbosity category exports."""
+
+from __future__ import annotations
+
+from easydiffraction.project.categories.verbosity.default import Verbosity
+from easydiffraction.project.categories.verbosity.factory import VerbosityFactory
diff --git a/src/easydiffraction/project/categories/verbosity/default.py b/src/easydiffraction/project/categories/verbosity/default.py
new file mode 100644
index 000000000..09b3821c6
--- /dev/null
+++ b/src/easydiffraction/project/categories/verbosity/default.py
@@ -0,0 +1,55 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Project fit-output verbosity category."""
+
+from __future__ import annotations
+
+from easydiffraction.core.category import CategoryItem
+from easydiffraction.core.metadata import TypeInfo
+from easydiffraction.core.validation import AttributeSpec
+from easydiffraction.core.validation import MembershipValidator
+from easydiffraction.core.variable import StringDescriptor
+from easydiffraction.io.cif.handler import CifHandler
+from easydiffraction.project.categories.verbosity.factory import VerbosityFactory
+from easydiffraction.utils.enums import VerbosityEnum
+
+
+@VerbosityFactory.register
+class Verbosity(CategoryItem):
+    """Fit-output verbosity selection for a project."""
+
+    _category_code = 'verbosity'
+
+    type_info = TypeInfo(
+        tag='default',
+        description='Project verbosity category',
+    )
+
+    def __init__(self) -> None:
+        super().__init__()
+
+        self._fit = StringDescriptor(
+            name='fit',
+            description='Fitting process output verbosity',
+            value_spec=AttributeSpec(
+                default=VerbosityEnum.default().value,
+                validator=MembershipValidator(
+                    allowed=[member.value for member in VerbosityEnum],
+                ),
+            ),
+            cif_handler=CifHandler(names=['_verbosity.fit']),
+        )
+
+    @property
+    def fit(self) -> StringDescriptor:
+        """Fitting process output verbosity."""
+        return self._fit
+
+    @fit.setter
+    def fit(self, value: str) -> None:
+        self._fit.value = VerbosityEnum(value).value
+
+    @property
+    def as_cif(self) -> str:
+        """Return CIF representation of this verbosity category."""
+        return super().as_cif
diff --git a/src/easydiffraction/project/categories/verbosity/factory.py b/src/easydiffraction/project/categories/verbosity/factory.py
new file mode 100644
index 000000000..28b26bb97
--- /dev/null
+++ b/src/easydiffraction/project/categories/verbosity/factory.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Factory for project verbosity categories."""
+
+from __future__ import annotations
+
+from typing import ClassVar
+
+from easydiffraction.core.factory import FactoryBase
+
+
+class VerbosityFactory(FactoryBase):
+    """Create project verbosity category instances."""
+
+    _default_rules: ClassVar[dict] = {
+        frozenset(): 'default',
+    }
diff --git a/src/easydiffraction/project/display.py b/src/easydiffraction/project/display.py
index 83a9e40f2..5b760d997 100644
--- a/src/easydiffraction/project/display.py
+++ b/src/easydiffraction/project/display.py
@@ -4,6 +4,7 @@
 
 from __future__ import annotations
 
+from contextlib import nullcontext
 from dataclasses import dataclass
 from typing import TYPE_CHECKING
 
@@ -134,6 +135,68 @@ class PosteriorDisplay:
     def __init__(self, project: Project) -> None:
         self._project = project
 
+    def _pairs_need_processing_indicator(
+        self,
+        *,
+        parameters: list[object] | None,
+    ) -> bool:
+        """
+        Return whether posterior pair plotting still needs processing.
+        """
+        if parameters is not None:
+            return True
+
+        analysis = self._project.analysis
+        sidecar_data = getattr(analysis, '_persisted_fit_state_sidecar', {})
+        pair_caches = sidecar_data.get('pair_caches', {})
+        return not (
+            analysis.bayesian_result.has_pair_cache.value
+            and len(analysis.bayesian_pair_caches) > 0
+            and bool(pair_caches)
+        )
+
+    def _predictive_needs_processing_indicator(
+        self,
+        *,
+        expt_name: str,
+        style: str,
+        x: object | None,
+    ) -> bool:
+        """Return whether predictive plotting still needs processing."""
+        analysis = self._project.analysis
+        sidecar_data = getattr(analysis, '_persisted_fit_state_sidecar', {})
+        predictive_datasets = sidecar_data.get('predictive_datasets', {})
+        if not (
+            analysis.bayesian_result.has_posterior_predictive.value
+            and bool(predictive_datasets)
+            and expt_name in predictive_datasets
+        ):
+            return True
+
+        experiment = self._project.experiments[expt_name]
+        plotter = self._project.rendering.plotter
+        _, x_axis_name, _, _, _ = plotter._resolve_x_axis(experiment.type, x)
+        require_draws = plotter.engine == PlotterEngineEnum.PLOTLY.value and style in {
+            'draws',
+            'band+draws',
+        }
+
+        matching_rows = [
+            row
+            for row in analysis.bayesian_predictive_datasets
+            if row.experiment_name.value == expt_name
+            and str(row.x_axis_name.value) == str(x_axis_name)
+        ]
+        if not matching_rows:
+            return True
+        if not require_draws:
+            return False
+        return not any(
+            row.draws_path.value is not None
+            and predictive_datasets[expt_name].get('draws') is not None
+            for row in matching_rows
+        )
+
     def pairs(
         self,
         parameters: list[object] | None = None,
@@ -143,10 +206,15 @@ def pairs(
         max_parameters: int = 6,
     ) -> None:
         """Plot posterior pair relationships for sampled parameters."""
-        with activity_indicator(
-            ACTIVITY_LABEL_PROCESSING,
-            verbosity=VerbosityEnum(self._project.verbosity),
-        ):
+        indicator_context = (
+            activity_indicator(
+                ACTIVITY_LABEL_PROCESSING,
+                verbosity=VerbosityEnum(self._project.verbosity.fit.value),
+            )
+            if self._pairs_need_processing_indicator(parameters=parameters)
+            else nullcontext()
+        )
+        with indicator_context:
             self._project.rendering.plotter.plot_posterior_pairs(
                 parameters=parameters,
                 style=style,
@@ -182,10 +250,19 @@ def predictive(
         x: object | None = None,
     ) -> None:
         """Plot posterior predictive summaries for one experiment."""
-        with activity_indicator(
-            ACTIVITY_LABEL_PROCESSING,
-            verbosity=VerbosityEnum(self._project.verbosity),
-        ):
+        indicator_context = (
+            activity_indicator(
+                ACTIVITY_LABEL_PROCESSING,
+                verbosity=VerbosityEnum(self._project.verbosity.fit.value),
+            )
+            if self._predictive_needs_processing_indicator(
+                expt_name=expt_name,
+                style=style,
+                x=x,
+            )
+            else nullcontext()
+        )
+        with indicator_context:
             self._project.rendering.plotter.plot_posterior_predictive(
                 expt_name=expt_name,
                 style=style,
@@ -249,10 +326,19 @@ def pattern(
                 msg = self._status_by_name(statuses, 'auto').reason
                 raise ValueError(msg)
             if 'uncertainty' in auto_include:
-                with activity_indicator(
-                    ACTIVITY_LABEL_PROCESSING,
-                    verbosity=VerbosityEnum(self._project.verbosity),
-                ):
+                indicator_context = (
+                    activity_indicator(
+                        ACTIVITY_LABEL_PROCESSING,
+                        verbosity=VerbosityEnum(self._project.verbosity.fit.value),
+                    )
+                    if self._posterior._predictive_needs_processing_indicator(
+                        expt_name=expt_name,
+                        style='band',
+                        x=x,
+                    )
+                    else nullcontext()
+                )
+                with indicator_context:
                     self._project.rendering.plotter._plot_posterior_predictive_request(
                         expt_name=expt_name,
                         style='band',
@@ -283,10 +369,19 @@ def pattern(
             raise ValueError(msg)
 
         if 'uncertainty' in normalized_include:
-            with activity_indicator(
-                ACTIVITY_LABEL_PROCESSING,
-                verbosity=VerbosityEnum(self._project.verbosity),
-            ):
+            indicator_context = (
+                activity_indicator(
+                    ACTIVITY_LABEL_PROCESSING,
+                    verbosity=VerbosityEnum(self._project.verbosity.fit.value),
+                )
+                if self._posterior._predictive_needs_processing_indicator(
+                    expt_name=expt_name,
+                    style='band',
+                    x=x,
+                )
+                else nullcontext()
+            )
+            with indicator_context:
                 self._project.rendering.plotter._plot_posterior_predictive_request(
                     expt_name=expt_name,
                     style='band',
@@ -739,12 +834,16 @@ def _uncertainty_status(
         if fit_results is None:
             return False, 'No fit results are available.'
 
-        posterior_samples = getattr(fit_results, 'posterior_samples', None)
         posterior_predictive = getattr(fit_results, 'posterior_predictive', None)
-        if posterior_samples is None or posterior_predictive is None:
+        if not posterior_predictive:
             return False, 'Posterior predictive data is unavailable.'
 
-        if self._project.rendering.chart_engine.value != PlotterEngineEnum.PLOTLY.value:
+        active_chart_engine = getattr(self._project.rendering.plotter, 'engine', None)
+        if active_chart_engine is None:
+            chart_engine = getattr(self._project.rendering, 'chart_engine', None)
+            active_chart_engine = getattr(chart_engine, 'value', None)
+
+        if active_chart_engine != PlotterEngineEnum.PLOTLY.value:
             return False, 'Uncertainty bands currently require the Plotly chart engine.'
 
         return True, ''
diff --git a/src/easydiffraction/project/project.py b/src/easydiffraction/project/project.py
index 5bbe8acba..9b89eb500 100644
--- a/src/easydiffraction/project/project.py
+++ b/src/easydiffraction/project/project.py
@@ -5,6 +5,7 @@
 from __future__ import annotations
 
 import pathlib
+import shutil
 import tempfile
 from typing import TYPE_CHECKING
 from typing import ClassVar
@@ -16,17 +17,25 @@
 from easydiffraction.core.guard import GuardedBase
 from easydiffraction.datablocks.experiment.collection import Experiments
 from easydiffraction.datablocks.structure.collection import Structures
+from easydiffraction.io.cif.serialize import analysis_from_cif
+from easydiffraction.io.cif.serialize import project_config_from_cif
 from easydiffraction.io.cif.serialize import project_config_to_cif
 from easydiffraction.io.cif.serialize import project_to_cif
+from easydiffraction.io.results_sidecar import read_analysis_results_sidecar
+from easydiffraction.io.results_sidecar import write_analysis_results_sidecar
 from easydiffraction.project.display import ProjectDisplay
 from easydiffraction.project.project_config import ProjectConfig
 from easydiffraction.summary.summary import Summary
 from easydiffraction.utils.enums import VerbosityEnum
+from easydiffraction.utils.environment import resolve_artifact_path
 from easydiffraction.utils.logging import console
 from easydiffraction.utils.logging import log
 
 if TYPE_CHECKING:
+    from collections.abc import Callable
+
     from easydiffraction.project.categories.rendering import Rendering
+    from easydiffraction.project.categories.verbosity import Verbosity
     from easydiffraction.project.project_info import ProjectInfo
 
 
@@ -108,6 +117,63 @@ def _resolve_data_path_from_results_csv(
     return project_path / path
 
 
+def _load_cif_directory(
+    cif_dir: pathlib.Path,
+    add_from_cif_path: Callable[[str], None],
+) -> None:
+    """Load all CIF files from one directory using the given loader."""
+    if not cif_dir.is_dir():
+        return
+
+    for cif_file in sorted(cif_dir.glob('*.cif')):
+        add_from_cif_path(str(cif_file))
+
+
+def _create_loading_project(project_cls: type[Project]) -> Project:
+    """Create a project instance while suppressing varname lookup."""
+    project_cls._loading = True
+    try:
+        return project_cls()
+    finally:
+        project_cls._loading = False
+
+
+def _load_project_info(project: Project, project_path: pathlib.Path) -> None:
+    """
+    Restore project configuration from ``project.cif`` when present.
+    """
+    project_cif_path = project_path / 'project.cif'
+    if project_cif_path.is_file():
+        project_config_from_cif(project, project_cif_path.read_text())
+
+
+def _resolved_analysis_cif_path(project_path: pathlib.Path) -> pathlib.Path | None:
+    """Return the preferred analysis CIF path for a saved project."""
+    analysis_cif_path = project_path / 'analysis' / 'analysis.cif'
+    if analysis_cif_path.is_file():
+        return analysis_cif_path
+
+    analysis_cif_path = project_path / 'analysis.cif'
+    if analysis_cif_path.is_file():
+        return analysis_cif_path
+    return None
+
+
+def _load_project_analysis(project: Project, project_path: pathlib.Path) -> None:
+    """Restore analysis categories and sidecar state from disk."""
+    analysis_cif_path = _resolved_analysis_cif_path(project_path)
+    if analysis_cif_path is None:
+        return
+
+    analysis_from_cif(project._analysis, analysis_cif_path.read_text())
+    read_analysis_results_sidecar(
+        analysis=project._analysis,
+        analysis_dir=analysis_cif_path.parent,
+    )
+    if project._analysis._has_persisted_fit_state():
+        project._analysis._restore_live_parameter_state(project._build_parameter_map())
+
+
 class Project(GuardedBase):
     """
     Central API for managing a diffraction data analysis project.
@@ -135,12 +201,12 @@ def __init__(
         self._structures = Structures()
         self._experiments = Experiments()
         object.__setattr__(self, '_rendering', self._config.rendering)
+        object.__setattr__(self, '_verbosity', self._config.verbosity)
         self._display = ProjectDisplay(self)
         self._analysis = Analysis(self)
         self._summary = Summary(self)
         self._saved = False
         self._varname = 'project' if type(self)._loading else varname()
-        self._verbosity: VerbosityEnum = VerbosityEnum.FULL
         type(self)._current_project = self
 
     @classmethod
@@ -249,21 +315,14 @@ def as_cif(self) -> str:
         return project_to_cif(self)
 
     @property
-    def verbosity(self) -> str:
-        """
-        Project-wide console output verbosity.
-
-        Returns
-        -------
-        str
-            One of ``'full'``, ``'short'``, or ``'silent'``.
-        """
-        return self._verbosity.value
+    def verbosity(self) -> Verbosity:
+        """Verbosity configuration bound to the project."""
+        return self._verbosity
 
     @verbosity.setter
     def verbosity(self, value: str) -> None:
         """
-        Set project-wide console output verbosity.
+        Set fitting process output verbosity.
 
         Parameters
         ----------
@@ -271,7 +330,7 @@ def verbosity(self, value: str) -> None:
             ``'full'`` for multi-line output, ``'short'`` for one-line
             status messages, or ``'silent'`` for no output.
         """
-        self._verbosity = VerbosityEnum(value)
+        self._verbosity.fit = VerbosityEnum(value).value
 
     # ------------------------------------------
     #  Project File I/O
@@ -303,52 +362,19 @@ def load(cls, dir_path: str) -> Project:
         FileNotFoundError
             If *dir_path* does not exist.
         """
-        from easydiffraction.io.cif.serialize import analysis_from_cif  # noqa: PLC0415
-        from easydiffraction.io.cif.serialize import project_config_from_cif  # noqa: PLC0415
-
         project_path = pathlib.Path(dir_path)
         if not project_path.is_dir():
             msg = f"Project directory not found: '{dir_path}'"
             raise FileNotFoundError(msg)
 
-        # Create a minimal project.
-        # Use _loading sentinel to skip varname() inside __init__.
-        cls._loading = True
-        try:
-            project = cls()
-        finally:
-            cls._loading = False
+        project = _create_loading_project(cls)
         project._saved = True
 
-        # 1. Load project info
-        project_cif_path = project_path / 'project.cif'
-        if project_cif_path.is_file():
-            cif_text = project_cif_path.read_text()
-            project_config_from_cif(project, cif_text)
-
+        _load_project_info(project, project_path)
         project.info.path = project_path
-
-        # 2. Load structures
-        structures_dir = project_path / 'structures'
-        if structures_dir.is_dir():
-            for cif_file in sorted(structures_dir.glob('*.cif')):
-                project._structures.add_from_cif_path(str(cif_file))
-
-        # 3. Load experiments
-        experiments_dir = project_path / 'experiments'
-        if experiments_dir.is_dir():
-            for cif_file in sorted(experiments_dir.glob('*.cif')):
-                project._experiments.add_from_cif_path(str(cif_file))
-
-        # 4. Load analysis
-        #    Check analysis/analysis.cif first (future layout), then
-        #    fall back to analysis.cif at root (current layout).
-        analysis_cif_path = project_path / 'analysis' / 'analysis.cif'
-        if not analysis_cif_path.is_file():
-            analysis_cif_path = project_path / 'analysis.cif'
-        if analysis_cif_path.is_file():
-            cif_text = analysis_cif_path.read_text()
-            analysis_from_cif(project._analysis, cif_text)
+        _load_cif_directory(project_path / 'structures', project._structures.add_from_cif_path)
+        _load_cif_directory(project_path / 'experiments', project._experiments.add_from_cif_path)
+        _load_project_analysis(project, project_path)
 
         # 5. Resolve alias param references
         project._resolve_alias_references()
@@ -373,13 +399,7 @@ def _resolve_alias_references(self) -> None:
         if not aliases._items:
             return
 
-        # Build unique_name → parameter map
-        all_params = self._structures.parameters + self._experiments.parameters
-        param_map: dict[str, object] = {}
-        for p in all_params:
-            uname = getattr(p, 'unique_name', None)
-            if uname is not None:
-                param_map[uname] = p
+        param_map = self._build_parameter_map()
 
         for alias in aliases:
             uname = alias.param_unique_name.value
@@ -391,6 +411,22 @@ def _resolve_alias_references(self) -> None:
                     f"parameter '{uname}'. Reference not resolved."
                 )
 
+    def _build_parameter_map(self) -> dict[str, object]:
+        """
+        Return a ``unique_name`` to live parameter mapping.
+
+        The map combines structure and experiment parameters and is
+        reused by CIF restore steps that need to reconnect persisted
+        names to live parameter objects.
+        """
+        all_params = self._structures.parameters + self._experiments.parameters
+        param_map: dict[str, object] = {}
+        for param in all_params:
+            unique_name = getattr(param, 'unique_name', None)
+            if unique_name is not None:
+                param_map[unique_name] = param
+        return param_map
+
     def save(self) -> None:
         """Save the project into the existing project directory."""
         if self.info.path is None:
@@ -441,6 +477,10 @@ def save(self) -> None:
         with (analysis_dir / 'analysis.cif').open('w') as f:
             f.write(self.analysis.as_cif)
             console.print('├── 📁 analysis/')
+        write_analysis_results_sidecar(
+            analysis=self.analysis,
+            analysis_dir=analysis_dir,
+        )
 
         analysis_file_names = sorted(
             path.name for path in analysis_dir.iterdir() if path.is_file()
@@ -462,12 +502,40 @@ def save_as(
         dir_path: str,
         *,
         temporary: bool = False,
+        overwrite: bool = True,
     ) -> None:
-        """Save the project into a new directory."""
+        """
+        Save the project into a directory.
+
+        Parameters
+        ----------
+        dir_path : str
+            Destination directory for the saved project.
+        temporary : bool, default=False
+            Whether to save beneath the system temporary directory.
+        overwrite : bool, default=True
+            Whether to remove an existing target directory before
+            saving.
+        """
         if temporary:
             tmp: str = tempfile.gettempdir()
-            dir_path = pathlib.Path(tmp) / dir_path
-        self.info.path = dir_path
+            project_dir = pathlib.Path(tmp) / dir_path
+        else:
+            project_dir = resolve_artifact_path(dir_path)
+
+        if overwrite and project_dir.is_dir():
+            current_working_directory = pathlib.Path.cwd().resolve()
+            resolved_project_dir = project_dir.resolve()
+            if resolved_project_dir == current_working_directory:
+                for child_path in resolved_project_dir.iterdir():
+                    if child_path.is_dir():
+                        shutil.rmtree(child_path)
+                    else:
+                        child_path.unlink()
+            else:
+                shutil.rmtree(project_dir)
+
+        self.info.path = project_dir
         self.save()
 
     def apply_params_from_csv(self, row_index: int) -> None:
diff --git a/src/easydiffraction/project/project_config.py b/src/easydiffraction/project/project_config.py
index 32147b609..dd92137ab 100644
--- a/src/easydiffraction/project/project_config.py
+++ b/src/easydiffraction/project/project_config.py
@@ -9,6 +9,8 @@
 from easydiffraction.project.categories.info import ProjectInfoFactory
 from easydiffraction.project.categories.rendering import Rendering
 from easydiffraction.project.categories.rendering import RenderingFactory
+from easydiffraction.project.categories.verbosity import Verbosity
+from easydiffraction.project.categories.verbosity import VerbosityFactory
 
 
 class ProjectConfig(CategoryOwner):
@@ -28,6 +30,7 @@ def __init__(
             description=description,
         )
         self._rendering = RenderingFactory.create(RenderingFactory.default_tag())
+        self._verbosity = VerbosityFactory.create(VerbosityFactory.default_tag())
 
     @property
     def info(self) -> ProjectInfo:
@@ -39,6 +42,11 @@ def rendering(self) -> Rendering:
         """Rendering configuration category."""
         return self._rendering
 
+    @property
+    def verbosity(self) -> Verbosity:
+        """Verbosity configuration category."""
+        return self._verbosity
+
     @property
     def as_cif(self) -> str:
         """Serialize singleton project categories to CIF."""
diff --git a/src/easydiffraction/utils/environment.py b/src/easydiffraction/utils/environment.py
index e2df97d48..0fd35ce0c 100644
--- a/src/easydiffraction/utils/environment.py
+++ b/src/easydiffraction/utils/environment.py
@@ -5,7 +5,55 @@
 
 import os
 import sys
+import tempfile
 from importlib.util import find_spec
+from pathlib import Path
+
+_ARTIFACT_ROOT_ENV_VAR = 'EASYDIFFRACTION_ARTIFACT_ROOT'
+_PIXI_PROJECT_ROOT_ENV_VAR = 'PIXI_PROJECT_ROOT'
+_TUTORIALS_DIR = Path('docs') / 'docs' / 'tutorials'
+_TUTORIAL_ARTIFACT_ROOT = Path('tmp') / 'tutorials'
+
+
+def _repo_root() -> Path | None:
+    project_root = os.environ.get(_PIXI_PROJECT_ROOT_ENV_VAR)
+    if project_root:
+        return Path(project_root).resolve()
+
+    for parent in Path(__file__).resolve().parents:
+        if (parent / 'pixi.toml').is_file() and (parent / _TUTORIALS_DIR).is_dir():
+            return parent
+
+    return None
+
+
+def _tutorial_artifact_root() -> Path | None:
+    repo_root = _repo_root()
+    if repo_root is None:
+        return None
+
+    tutorials_dir = (repo_root / _TUTORIALS_DIR).resolve()
+    cwd = Path.cwd().resolve()
+    if not cwd.is_relative_to(tutorials_dir):
+        return None
+
+    return (repo_root / _TUTORIAL_ARTIFACT_ROOT).resolve()
+
+
+def _artifact_root() -> Path | None:
+    artifact_root = os.environ.get(_ARTIFACT_ROOT_ENV_VAR)
+    if not artifact_root:
+        return _tutorial_artifact_root()
+
+    root = Path(artifact_root)
+    if root.is_absolute():
+        return root.resolve()
+
+    project_root = os.environ.get(_PIXI_PROJECT_ROOT_ENV_VAR)
+    if project_root:
+        return (Path(project_root) / root).resolve()
+
+    return (Path.cwd() / root).resolve()
 
 
 def in_pytest() -> bool:
@@ -107,6 +155,52 @@ def in_github_ci() -> bool:
     return os.environ.get('GITHUB_ACTIONS') is not None
 
 
+def resolve_artifact_path(path: str | Path) -> Path:
+    """
+    Resolve a path against the configured artifact root.
+
+    Parameters
+    ----------
+    path : str | Path
+        Path to resolve.
+
+    Returns
+    -------
+    Path
+        The original path when no artifact root is configured or when
+        *path* is absolute. Otherwise, the absolute path under the
+        configured artifact root.
+    """
+    resolved_path = Path(path)
+    artifact_root = _artifact_root()
+    if artifact_root is None or resolved_path.is_absolute():
+        return resolved_path
+
+    return (artifact_root / resolved_path).resolve()
+
+
+def create_artifact_temp_dir(prefix: str) -> Path:
+    """
+    Create a temporary directory under the artifact root when set.
+
+    Parameters
+    ----------
+    prefix : str
+        Prefix for the temporary directory name.
+
+    Returns
+    -------
+    Path
+        Path to the created temporary directory.
+    """
+    artifact_root = _artifact_root()
+    if artifact_root is None:
+        return Path(tempfile.mkdtemp(prefix=prefix))
+
+    artifact_root.mkdir(parents=True, exist_ok=True)
+    return Path(tempfile.mkdtemp(prefix=prefix, dir=artifact_root)).resolve()
+
+
 # ----------------------------------------------------------------------
 # IPython/Jupyter helpers
 # ----------------------------------------------------------------------
diff --git a/src/easydiffraction/utils/utils.py b/src/easydiffraction/utils/utils.py
index cf461aeea..6dba4301d 100644
--- a/src/easydiffraction/utils/utils.py
+++ b/src/easydiffraction/utils/utils.py
@@ -6,6 +6,7 @@
 import functools
 import json
 import pathlib
+import shutil
 import urllib.request
 from importlib.metadata import PackageNotFoundError
 from importlib.metadata import version
@@ -20,6 +21,8 @@
 from uncertainties import ufloat_fromstr
 
 from easydiffraction.display.tables import TableRenderer
+from easydiffraction.io.ascii import extract_project_from_zip
+from easydiffraction.utils.environment import resolve_artifact_path
 from easydiffraction.utils.logging import console
 from easydiffraction.utils.logging import log
 
@@ -28,9 +31,9 @@
 _DATA_REPO = 'easyscience/diffraction'
 _DATA_ROOT = 'data'
 # commit SHA preferred
-_DATA_INDEX_REF = '39dad256ba1faedf4b26fad3e44a361c802fd8e4'
+_DATA_INDEX_REF = 'dbe92a87e0106c4742eee0ff9a8e32bdb8b483cb'
 # macOS: sha256sum index.json
-_DATA_INDEX_HASH = 'sha256:301aaca0f35927cd63715b858a1f03164e4d05d1d39234325a3798d2b4a5f4ea'
+_DATA_INDEX_HASH = 'sha256:9e7bbaf2cb650f4126572e85157c63bc76f201408856fe4af566bee55dcdfbb4'
 
 
 def _build_data_url(path: str) -> str:
@@ -112,6 +115,42 @@ def _fetch_data_index() -> dict:
         return json.load(f)
 
 
+def _existing_project_dir(extraction_dir: pathlib.Path) -> pathlib.Path | None:
+    """Return one extracted project directory from a destination."""
+    project_files = sorted(extraction_dir.rglob('project.cif'))
+    if not project_files:
+        return None
+    return project_files[0].parent.resolve()
+
+
+def _download_data_message(data_id: int | str, record: dict) -> str:
+    """Return the console message for one downloadable data record."""
+    description = record.get('description', '')
+    message = f'Data #{data_id}'
+    if description:
+        message += f': {description}'
+    return message
+
+
+def _download_data_targets(
+    data_id: int | str,
+    destination: str,
+    record: dict,
+) -> tuple[str, bool, pathlib.Path, pathlib.Path, pathlib.Path, str]:
+    """Return URL and filesystem targets for one download request."""
+    record_path = _record_path(record)
+    url = _build_data_url(record_path)
+    _validate_url(url)
+
+    fname = _filename_for_id_from_path(data_id, record_path)
+    is_project_archive = record.get('kind') == 'project' and fname.endswith('.zip')
+    dest_path = resolve_artifact_path(destination)
+    dest_path.mkdir(parents=True, exist_ok=True)
+    file_path = dest_path / fname
+    extraction_dir = dest_path / pathlib.Path(fname).stem
+    return url, is_project_archive, dest_path, file_path, extraction_dir, fname
+
+
 @functools.lru_cache(maxsize=1)
 def _fetch_tutorials_index() -> dict:
     """
@@ -160,14 +199,18 @@ def download_data(
     id : int | str
         Numeric dataset id (e.g. 12).
     destination : str, default='data'
-        Directory to save the file into (created if missing).
+        Directory to save the downloaded file or extracted project into
+        (created if missing). Relative destinations are resolved against
+        the configured artifact root when
+        ``EASYDIFFRACTION_ARTIFACT_ROOT`` is set.
     overwrite : bool, default=False
         Whether to overwrite the file if it already exists.
 
     Returns
     -------
     str
-        Full path to the downloaded file as string.
+        Full path to the downloaded file, or to the extracted project
+        directory for project ZIP archives, as string.
 
     Raises
     ------
@@ -186,24 +229,29 @@ def download_data(
         raise KeyError(msg)
 
     record = index[key]
-    record_path = _record_path(record)
-    url = _build_data_url(record_path)
-    _validate_url(url)
-    fname = _filename_for_id_from_path(id, record_path)
-
-    dest_path = pathlib.Path(destination)
-    dest_path.mkdir(parents=True, exist_ok=True)
-    file_path = dest_path / fname
-
-    description = record.get('description', '')
-    message = f'Data #{id}'
-    if description:
-        message += f': {description}'
+    url, is_project_archive, dest_path, file_path, extraction_dir, fname = _download_data_targets(
+        id, destination, record
+    )
+    message = _download_data_message(id, record)
 
     console.paragraph('Getting data...')
     console.print(f'{message}')
 
+    if is_project_archive and extraction_dir.exists() and not overwrite:
+        existing_project_dir = _existing_project_dir(extraction_dir)
+        if existing_project_dir is not None:
+            console.print(
+                f"✅ Data #{id} already extracted at '{existing_project_dir}'. "
+                'Keeping existing project.'
+            )
+            return str(existing_project_dir)
+
     if file_path.exists():
+        if is_project_archive and not overwrite:
+            project_dir = extract_project_from_zip(file_path, destination=extraction_dir)
+            file_path.unlink()
+            console.print(f"✅ Data #{id} extracted to '{project_dir}'")
+            return str(project_dir)
         if not overwrite:
             console.print(
                 f"✅ Data #{id} already present at '{file_path}'. Keeping existing file."
@@ -214,6 +262,9 @@ def download_data(
 
     known_hash = _normalize_known_hash(record.get('hash'))
 
+    if is_project_archive and extraction_dir.exists() and overwrite:
+        shutil.rmtree(extraction_dir)
+
     # Pooch downloads to destination with our controlled filename.
     pooch.retrieve(
         url=url,
@@ -222,10 +273,45 @@ def download_data(
         path=str(dest_path),
     )
 
-    console.print(f"✅ Data #{id} downloaded to '{file_path}'")
+    if is_project_archive:
+        project_dir = extract_project_from_zip(file_path, destination=extraction_dir)
+        file_path.unlink()
+        console.print(f"✅ Data #{id} downloaded and extracted to\n'{project_dir}'")
+        return str(project_dir)
+
+    console.print(f"✅ Data #{id} downloaded to:\n'{file_path}'")
     return str(file_path)
 
 
+def list_data() -> None:
+    """Display a table of available example data records."""
+    index = _fetch_data_index()
+    if not index:
+        console.print('❌ No example data available.')
+        return
+
+    console.paragraph('Example data available for download:')
+
+    columns_headers = ['id', 'file', 'kind', 'description']
+    columns_alignment = ['right', 'left', 'left', 'left']
+    columns_data = []
+
+    for data_id in sorted(index, key=lambda value: int(value) if value.isdigit() else value):
+        record = index[data_id]
+        columns_data.append([
+            data_id,
+            pathlib.PurePosixPath(_record_path(record)).name,
+            record.get('kind', ''),
+            record.get('description', ''),
+        ])
+
+    render_table(
+        columns_headers=columns_headers,
+        columns_data=columns_data,
+        columns_alignment=columns_alignment,
+    )
+
+
 def package_version(package_name: str) -> str | None:
     """
     Get the installed version string of the specified package.
@@ -474,7 +560,7 @@ def download_tutorial(
     with _safe_urlopen(url) as resp:
         file_path.write_bytes(resp.read())
 
-    console.print(f"✅ Tutorial #{id} downloaded to '{file_path}'")
+    console.print(f"✅ Tutorial #{id} downloaded to:\n'{file_path}'")
     return str(file_path)
 
 
diff --git a/tests/functional/test_project_lifecycle.py b/tests/functional/test_project_lifecycle.py
index 7d9c5a190..850f040e9 100644
--- a/tests/functional/test_project_lifecycle.py
+++ b/tests/functional/test_project_lifecycle.py
@@ -88,7 +88,7 @@ def test_default_verbosity_is_full(self):
             project = Project()
         finally:
             Project._loading = False
-        assert project.verbosity == 'full'
+        assert project.verbosity.fit.value == 'full'
 
     def test_set_verbosity_short(self):
         Project._loading = True
@@ -97,7 +97,7 @@ def test_set_verbosity_short(self):
         finally:
             Project._loading = False
         project.verbosity = 'short'
-        assert project.verbosity == 'short'
+        assert project.verbosity.fit.value == 'short'
 
     def test_set_verbosity_silent(self):
         Project._loading = True
@@ -106,7 +106,7 @@ def test_set_verbosity_silent(self):
         finally:
             Project._loading = False
         project.verbosity = 'silent'
-        assert project.verbosity == 'silent'
+        assert project.verbosity.fit.value == 'silent'
 
     def test_invalid_verbosity_raises(self):
         Project._loading = True
diff --git a/tests/integration/fitting/test_bayesian_dream.py b/tests/integration/fitting/test_bayesian_dream.py
index d6695973b..ef52c1f1a 100644
--- a/tests/integration/fitting/test_bayesian_dream.py
+++ b/tests/integration/fitting/test_bayesian_dream.py
@@ -163,7 +163,7 @@ def test_lm_prefit_followed_by_dream_uses_uncertainty_based_bounds():
     assert len(results.posterior_parameter_summaries) == 3
 
 
-def test_bayesian_fit_results_are_runtime_only_after_save_load(tmp_path):
+def test_bayesian_fit_results_reload_from_persisted_fit_state(tmp_path):
     project = _create_lbco_project()
     length_a, scale, offset = _dream_parameters(project)
     for parameter in (length_a, scale, offset):
@@ -185,8 +185,13 @@ def test_bayesian_fit_results_are_runtime_only_after_save_load(tmp_path):
     project.save_as(str(proj_dir))
 
     analysis_cif = proj_dir / 'analysis' / 'analysis.cif'
+    results_sidecar = proj_dir / 'analysis' / 'results.h5'
     assert analysis_cif.is_file()
-    assert 'posterior' not in analysis_cif.read_text().lower()
+    assert results_sidecar.is_file()
 
     loaded = Project.load(str(proj_dir))
-    assert loaded.analysis.fit_results is None
+    loaded_results = loaded.analysis.fit_results
+    assert loaded_results is not None
+    assert loaded_results.sampler_completed is True
+    assert loaded_results.posterior_samples is not None
+    assert loaded_results.posterior_samples.parameter_samples.ndim == 3
diff --git a/tests/integration/fitting/test_bumps_dream_support.py b/tests/integration/fitting/test_bumps_dream_support.py
index c09f0b92e..a824347e1 100644
--- a/tests/integration/fitting/test_bumps_dream_support.py
+++ b/tests/integration/fitting/test_bumps_dream_support.py
@@ -277,10 +277,7 @@ def test_build_mapper_falls_back_for_serial_and_unpicklable(monkeypatch):
     monkeypatch.setattr(
         'easydiffraction.analysis.minimizers.bumps_dream.can_pickle', lambda problem: False
     )
-    monkeypatch.setattr(
-        'easydiffraction.analysis.minimizers.bumps_dream.log.warning',
-        lambda message: warnings.append(message),
-    )
+    monkeypatch.setattr(minimizer, '_warn_after_tracking', warnings.append)
 
     assert minimizer._build_mapper('problem') is None
     assert warnings == [
diff --git a/tests/integration/fitting/test_sequential.py b/tests/integration/fitting/test_sequential.py
index da4c28774..73c068b3b 100644
--- a/tests/integration/fitting/test_sequential.py
+++ b/tests/integration/fitting/test_sequential.py
@@ -165,9 +165,18 @@ def test_fit_sequential_produces_csv(tmp_path) -> None:
 
     assert len(rows) == 3, f'Expected 3 rows, got {len(rows)}'
 
-    # Each row should have fit_success
+    # Each row should have fit_result.success
     for row in rows:
-        assert row['fit_success'] == 'True', f'Fit failed for {row["file_path"]}'
+        assert row['fit_result.success'] == 'True', f'Fit failed for {row["file_path"]}'
+        assert int(row['fit_result.iterations']) > 0, (
+            f'Expected non-zero iterations for {row["file_path"]}'
+        )
+
+    assert 'fit_result.reduced_chi_square' in rows[0]
+    assert 'fit_result.iterations' in rows[0]
+    assert 'success' not in rows[0]
+    assert 'reduced_chi_square' not in rows[0]
+    assert 'iterations' not in rows[0]
 
     # Each row should have parameter values
     assert 'lbco.cell.length_a' in rows[0]
@@ -316,7 +325,10 @@ def test_fit_sequential_parallel(tmp_path) -> None:
     assert len(rows) == 3, f'Expected 3 rows, got {len(rows)}'
 
     for row in rows:
-        assert row['fit_success'] == 'True', f'Fit failed for {row["file_path"]}'
+        assert row['fit_result.success'] == 'True', f'Fit failed for {row["file_path"]}'
+        assert int(row['fit_result.iterations']) > 0, (
+            f'Expected non-zero iterations for {row["file_path"]}'
+        )
 
     # Parameter values should be present and reasonable
     assert 'lbco.cell.length_a' in rows[0]
diff --git a/tests/unit/easydiffraction/analysis/categories/test_bayesian_convergence.py b/tests/unit/easydiffraction/analysis/categories/test_bayesian_convergence.py
new file mode 100644
index 000000000..dbcaf7299
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_bayesian_convergence.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/bayesian_convergence/."""
+
+
+def test_bayesian_convergence_factory_create():
+    from easydiffraction.analysis.categories.bayesian_convergence.default import (
+        BayesianConvergence,
+    )
+    from easydiffraction.analysis.categories.bayesian_convergence.factory import (
+        BayesianConvergenceFactory,
+    )
+
+    convergence = BayesianConvergenceFactory.create('default')
+
+    assert BayesianConvergenceFactory.default_tag() == 'default'
+    assert isinstance(convergence, BayesianConvergence)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_bayesian_distribution_caches.py b/tests/unit/easydiffraction/analysis/categories/test_bayesian_distribution_caches.py
new file mode 100644
index 000000000..26e91f732
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_bayesian_distribution_caches.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/bayesian_distribution_caches/."""
+
+
+def test_bayesian_distribution_caches_factory_create():
+    from easydiffraction.analysis.categories.bayesian_distribution_caches.default import (
+        BayesianDistributionCaches,
+    )
+    from easydiffraction.analysis.categories.bayesian_distribution_caches.factory import (
+        BayesianDistributionCachesFactory,
+    )
+
+    caches = BayesianDistributionCachesFactory.create('default')
+
+    assert BayesianDistributionCachesFactory.default_tag() == 'default'
+    assert isinstance(caches, BayesianDistributionCaches)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_bayesian_pair_caches.py b/tests/unit/easydiffraction/analysis/categories/test_bayesian_pair_caches.py
new file mode 100644
index 000000000..f4307f2c4
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_bayesian_pair_caches.py
@@ -0,0 +1,15 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/bayesian_pair_caches/."""
+
+
+def test_bayesian_pair_caches_factory_create():
+    from easydiffraction.analysis.categories.bayesian_pair_caches.default import BayesianPairCaches
+    from easydiffraction.analysis.categories.bayesian_pair_caches.factory import (
+        BayesianPairCachesFactory,
+    )
+
+    caches = BayesianPairCachesFactory.create('default')
+
+    assert BayesianPairCachesFactory.default_tag() == 'default'
+    assert isinstance(caches, BayesianPairCaches)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_bayesian_parameter_posteriors.py b/tests/unit/easydiffraction/analysis/categories/test_bayesian_parameter_posteriors.py
new file mode 100644
index 000000000..18aa5f019
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_bayesian_parameter_posteriors.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/bayesian_parameter_posteriors/."""
+
+
+def test_bayesian_parameter_posteriors_factory_create():
+    from easydiffraction.analysis.categories.bayesian_parameter_posteriors.default import (
+        BayesianParameterPosteriors,
+    )
+    from easydiffraction.analysis.categories.bayesian_parameter_posteriors.factory import (
+        BayesianParameterPosteriorsFactory,
+    )
+
+    posteriors = BayesianParameterPosteriorsFactory.create('default')
+
+    assert BayesianParameterPosteriorsFactory.default_tag() == 'default'
+    assert isinstance(posteriors, BayesianParameterPosteriors)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_bayesian_predictive_datasets.py b/tests/unit/easydiffraction/analysis/categories/test_bayesian_predictive_datasets.py
new file mode 100644
index 000000000..93746fb41
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_bayesian_predictive_datasets.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/bayesian_predictive_datasets/."""
+
+
+def test_bayesian_predictive_datasets_factory_create():
+    from easydiffraction.analysis.categories.bayesian_predictive_datasets.default import (
+        BayesianPredictiveDatasets,
+    )
+    from easydiffraction.analysis.categories.bayesian_predictive_datasets.factory import (
+        BayesianPredictiveDatasetsFactory,
+    )
+
+    datasets = BayesianPredictiveDatasetsFactory.create('default')
+
+    assert BayesianPredictiveDatasetsFactory.default_tag() == 'default'
+    assert isinstance(datasets, BayesianPredictiveDatasets)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_bayesian_result.py b/tests/unit/easydiffraction/analysis/categories/test_bayesian_result.py
new file mode 100644
index 000000000..e1d985385
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_bayesian_result.py
@@ -0,0 +1,13 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/bayesian_result/."""
+
+
+def test_bayesian_result_factory_create():
+    from easydiffraction.analysis.categories.bayesian_result.default import BayesianResult
+    from easydiffraction.analysis.categories.bayesian_result.factory import BayesianResultFactory
+
+    result = BayesianResultFactory.create('default')
+
+    assert BayesianResultFactory.default_tag() == 'default'
+    assert isinstance(result, BayesianResult)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_bayesian_sampler.py b/tests/unit/easydiffraction/analysis/categories/test_bayesian_sampler.py
new file mode 100644
index 000000000..976341765
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_bayesian_sampler.py
@@ -0,0 +1,13 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/bayesian_sampler/."""
+
+
+def test_bayesian_sampler_factory_create():
+    from easydiffraction.analysis.categories.bayesian_sampler.default import BayesianSampler
+    from easydiffraction.analysis.categories.bayesian_sampler.factory import BayesianSamplerFactory
+
+    sampler = BayesianSamplerFactory.create('default')
+
+    assert BayesianSamplerFactory.default_tag() == 'default'
+    assert isinstance(sampler, BayesianSampler)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_constraints.py b/tests/unit/easydiffraction/analysis/categories/test_constraints.py
index b887408ea..07016b298 100644
--- a/tests/unit/easydiffraction/analysis/categories/test_constraints.py
+++ b/tests/unit/easydiffraction/analysis/categories/test_constraints.py
@@ -22,6 +22,17 @@ def test_constraint_creation_and_collection():
     assert coll['a'].rhs_expr == 'b + c'
 
 
+def test_constraints_create_uses_explicit_id():
+    coll = Constraints()
+
+    coll.create(id='constraint_1', expression='a = b + c')
+
+    assert coll.names == ['constraint_1']
+    assert coll['constraint_1'].id.value == 'constraint_1'
+    assert coll['constraint_1'].lhs_alias == 'a'
+    assert coll['constraint_1'].rhs_expr == 'b + c'
+
+
 def test_constraints_from_cif_preserves_explicit_id_keys():
     doc = gemmi.cif.read_string(
         'data_constraints\n\n'
diff --git a/tests/unit/easydiffraction/analysis/categories/test_deterministic_result.py b/tests/unit/easydiffraction/analysis/categories/test_deterministic_result.py
new file mode 100644
index 000000000..ed764146d
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_deterministic_result.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/deterministic_result/."""
+
+
+def test_deterministic_result_factory_create():
+    from easydiffraction.analysis.categories.deterministic_result.default import (
+        DeterministicResult,
+    )
+    from easydiffraction.analysis.categories.deterministic_result.factory import (
+        DeterministicResultFactory,
+    )
+
+    result = DeterministicResultFactory.create('default')
+
+    assert DeterministicResultFactory.default_tag() == 'default'
+    assert isinstance(result, DeterministicResult)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_fit_parameter_correlations.py b/tests/unit/easydiffraction/analysis/categories/test_fit_parameter_correlations.py
new file mode 100644
index 000000000..8cb72d57b
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_fit_parameter_correlations.py
@@ -0,0 +1,17 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/fit_parameter_correlations/."""
+
+
+def test_fit_parameter_correlations_factory_create():
+    from easydiffraction.analysis.categories.fit_parameter_correlations.default import (
+        FitParameterCorrelations,
+    )
+    from easydiffraction.analysis.categories.fit_parameter_correlations.factory import (
+        FitParameterCorrelationsFactory,
+    )
+
+    correlations = FitParameterCorrelationsFactory.create('default')
+
+    assert FitParameterCorrelationsFactory.default_tag() == 'default'
+    assert isinstance(correlations, FitParameterCorrelations)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_fit_parameters.py b/tests/unit/easydiffraction/analysis/categories/test_fit_parameters.py
new file mode 100644
index 000000000..e3d0cedde
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_fit_parameters.py
@@ -0,0 +1,13 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/fit_parameters/."""
+
+
+def test_fit_parameters_factory_create():
+    from easydiffraction.analysis.categories.fit_parameters.default import FitParameters
+    from easydiffraction.analysis.categories.fit_parameters.factory import FitParametersFactory
+
+    collection = FitParametersFactory.create('default')
+
+    assert FitParametersFactory.default_tag() == 'default'
+    assert isinstance(collection, FitParameters)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_fit_result.py b/tests/unit/easydiffraction/analysis/categories/test_fit_result.py
new file mode 100644
index 000000000..4c9d869dc
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_fit_result.py
@@ -0,0 +1,13 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for analysis/categories/fit_result/."""
+
+
+def test_fit_result_factory_create():
+    from easydiffraction.analysis.categories.fit_result.default import FitResult
+    from easydiffraction.analysis.categories.fit_result.factory import FitResultFactory
+
+    fit_result = FitResultFactory.create('default')
+
+    assert FitResultFactory.default_tag() == 'default'
+    assert isinstance(fit_result, FitResult)
diff --git a/tests/unit/easydiffraction/analysis/categories/test_fit_state.py b/tests/unit/easydiffraction/analysis/categories/test_fit_state.py
new file mode 100644
index 000000000..236c798a5
--- /dev/null
+++ b/tests/unit/easydiffraction/analysis/categories/test_fit_state.py
@@ -0,0 +1,207 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for persisted fit-state analysis categories."""
+
+from __future__ import annotations
+
+import gemmi
+
+
+def test_fit_parameter_collection_is_empty_until_rows_exist():
+    from easydiffraction.analysis.categories.fit_parameters.default import FitParameters
+
+    collection = FitParameters()
+
+    assert collection.as_cif == ''
+
+
+def test_fit_parameter_collection_serializes_expected_tags_and_values():
+    from easydiffraction.analysis.categories.fit_parameters.default import FitParameters
+
+    collection = FitParameters()
+    collection.create(
+        param_unique_name='lbco.cell.length_a',
+        fit_min=3.88,
+        fit_max=3.90,
+        fit_bounds_uncertainty_multiplier=4.0,
+        start_value=3.89,
+        start_uncertainty=0.01,
+    )
+
+    cif_text = collection.as_cif
+
+    assert '_fit_parameter.param_unique_name' in cif_text
+    assert '_fit_parameter.fit_bounds_uncertainty_multiplier' in cif_text
+    assert 'lbco.cell.length_a' in cif_text
+
+
+def test_fit_result_serializes_expected_tags_and_enum_value():
+    from easydiffraction.analysis.categories.fit_result.default import FitResult
+
+    fit_result = FitResult()
+    fit_result._set_result_kind('bayesian')
+    fit_result._set_success(value=True)
+    fit_result._set_message('Sampler completed')
+    fit_result._set_iterations(3000)
+    fit_result._set_fitting_time(82.4)
+    fit_result._set_reduced_chi_square(1.031)
+
+    cif_text = fit_result.as_cif
+
+    assert '_fit_result.result_kind bayesian' in cif_text
+    assert '_fit_result.iterations 3000' in cif_text
+    assert '_fit_result.reduced_chi_square' in cif_text
+
+
+def test_fit_parameter_correlations_normalize_pair_order_and_replace_duplicate_ids():
+    from easydiffraction.analysis.categories.fit_parameter_correlations.default import (
+        FitParameterCorrelations,
+    )
+
+    correlations = FitParameterCorrelations()
+    correlations.create(
+        source_kind='posterior',
+        param_unique_name_i='z.param',
+        param_unique_name_j='a.param',
+        correlation=0.87,
+        id='1',
+    )
+    correlations.create(
+        source_kind='posterior',
+        param_unique_name_i='b.param',
+        param_unique_name_j='c.param',
+        correlation=0.55,
+        id='1',
+    )
+
+    assert len(correlations) == 1
+    assert correlations['1'].param_unique_name_i.value == 'b.param'
+    assert correlations['1'].param_unique_name_j.value == 'c.param'
+
+
+def test_fit_parameter_correlations_rebuild_index_from_cif():
+    from easydiffraction.analysis.categories.fit_parameter_correlations.default import (
+        FitParameterCorrelations,
+    )
+
+    cif_text = """data_fit_state
+loop_
+_fit_parameter_correlation.id
+_fit_parameter_correlation.source_kind
+_fit_parameter_correlation.param_unique_name_i
+_fit_parameter_correlation.param_unique_name_j
+_fit_parameter_correlation.correlation
+2 posterior hrpt.scale lbco.cell.length_a 0.42
+"""
+    document = gemmi.cif.read_string(cif_text)
+
+    correlations = FitParameterCorrelations()
+    correlations.from_cif(document.sole_block())
+
+    assert correlations.names == ['2']
+    assert correlations['2'].correlation.value == 0.42
+
+
+def test_bayesian_cache_manifest_collections_serialize_expected_keys():
+    from easydiffraction.analysis.categories.bayesian_distribution_caches.default import (
+        BayesianDistributionCaches,
+    )
+    from easydiffraction.analysis.categories.bayesian_pair_caches.default import (
+        BayesianPairCachePaths,
+        BayesianPairCaches,
+    )
+    from easydiffraction.analysis.categories.bayesian_predictive_datasets.default import (
+        BayesianPredictiveDatasetPaths,
+        BayesianPredictiveDatasets,
+    )
+
+    distributions = BayesianDistributionCaches()
+    distributions.create(
+        param_unique_name='lbco.cell.length_a',
+        x_path='/posterior/distribution/0/x',
+        density_path='/posterior/distribution/0/density',
+        n_grid=256,
+        n_draws_cached=48000,
+    )
+    pairs = BayesianPairCaches()
+    pairs.create(
+        parameter_names=('z.param', 'a.param'),
+        paths=BayesianPairCachePaths(
+            x_path='/posterior/pairs/0/x',
+            y_path='/posterior/pairs/0/y',
+            density_path='/posterior/pairs/0/density',
+            contour_level_path='/posterior/pairs/0/contour_levels',
+        ),
+        grid_shape=(64, 64),
+        n_draws_cached=4000,
+        id='7',
+    )
+    predictive = BayesianPredictiveDatasets()
+    predictive.create(
+        experiment_name='hrpt',
+        x_axis_name='two_theta',
+        paths=BayesianPredictiveDatasetPaths(
+            x_path='/predictive/hrpt/x',
+            best_sample_prediction_path='/predictive/hrpt/best_sample_prediction',
+            lower_95_path='/predictive/hrpt/lower_95',
+            upper_95_path='/predictive/hrpt/upper_95',
+        ),
+        n_x=2500,
+        n_draws_cached=0,
+    )
+
+    assert '_bayesian_distribution_cache.param_unique_name' in distributions.as_cif
+    assert pairs['7'].param_unique_name_x.value == 'a.param'
+    assert pairs['7'].param_unique_name_y.value == 'z.param'
+    assert '_bayesian_predictive_dataset.experiment_name' in predictive.as_cif
+
+
+def test_bayesian_sampler_and_convergence_use_integer_fields_in_cif():
+    from easydiffraction.analysis.categories.bayesian_convergence.default import (
+        BayesianConvergence,
+    )
+    from easydiffraction.analysis.categories.bayesian_sampler.default import BayesianSampler
+
+    sampler = BayesianSampler()
+    sampler._set_steps(100)
+    sampler._set_burn(20)
+    sampler._set_parallel(0)
+    sampler._set_random_seed(123)
+
+    convergence = BayesianConvergence()
+    convergence._set_n_draws(80)
+    convergence._set_n_chains(4)
+    convergence._set_n_parameters(3)
+
+    assert '_bayesian_sampler.steps 100' in sampler.as_cif
+    assert '_bayesian_sampler.parallel 0' in sampler.as_cif
+    assert '_bayesian_convergence.n_draws 80' in convergence.as_cif
+
+
+def test_bayesian_parameter_posteriors_preserve_row_order_from_cif():
+    from easydiffraction.analysis.categories.bayesian_parameter_posteriors.default import (
+        BayesianParameterPosteriors,
+    )
+
+    cif_text = """data_fit_state
+loop_
+_bayesian_parameter_posterior.unique_name
+_bayesian_parameter_posterior.display_name
+_bayesian_parameter_posterior.best_sample_value
+_bayesian_parameter_posterior.median
+_bayesian_parameter_posterior.uncertainty
+_bayesian_parameter_posterior.interval_68_lower
+_bayesian_parameter_posterior.interval_68_upper
+_bayesian_parameter_posterior.interval_95_lower
+_bayesian_parameter_posterior.interval_95_upper
+_bayesian_parameter_posterior.ess_bulk
+_bayesian_parameter_posterior.r_hat
+second.param second 2.0 2.1 0.2 1.9 2.3 1.8 2.4 20 1.01
+first.param first 1.0 1.1 0.1 0.9 1.3 0.8 1.4 10 1.00
+"""
+    document = gemmi.cif.read_string(cif_text)
+
+    posteriors = BayesianParameterPosteriors()
+    posteriors.from_cif(document.sole_block())
+
+    assert [row.unique_name.value for row in posteriors] == ['second.param', 'first.param']
diff --git a/tests/unit/easydiffraction/analysis/fit_helpers/test_tracking.py b/tests/unit/easydiffraction/analysis/fit_helpers/test_tracking.py
index 877a46b1a..ff9df3288 100644
--- a/tests/unit/easydiffraction/analysis/fit_helpers/test_tracking.py
+++ b/tests/unit/easydiffraction/analysis/fit_helpers/test_tracking.py
@@ -77,3 +77,50 @@ def test_tracker_fit_adds_timed_rows_and_resets_counter(monkeypatch):
     ]
     assert tracker._last_progress_time == 12.2
     assert tracker._previous_chi2 == 4.0
+
+
+def test_tracker_fit_progress_uses_backend_iterations_for_display():
+    from easydiffraction.analysis.fit_helpers.tracking import FitProgressTracker
+
+    tracker = FitProgressTracker()
+
+    tracker.track_fit_progress(iteration=1, reduced_chi2=10.0, elapsed_time=0.1)
+    tracker.track_fit_progress(iteration=63, reduced_chi2=5.0, elapsed_time=1.0)
+    tracker.track_fit_progress(iteration=122, reduced_chi2=4.0, elapsed_time=2.0)
+
+    assert tracker._df_rows == [
+        ['1', '0.10', '10.00', ''],
+        ['63', '1.00', '5.00', '50.0% ↓'],
+        ['122', '2.00', '4.00', '20.0% ↓'],
+    ]
+    assert tracker.best_iteration == 122
+
+
+def test_tracker_sampler_post_processing_adds_final_status_row():
+    from easydiffraction.analysis.fit_helpers.tracking import FitProgressTracker
+    from easydiffraction.analysis.fit_helpers.tracking import SamplerProgressUpdate
+
+    tracker = FitProgressTracker()
+    tracker.start_tracking('dream', mode='sampling')
+    tracker.start_timer()
+    tracker.track_sampler_progress(
+        SamplerProgressUpdate(
+            iteration=10,
+            total_iterations=10,
+            phase='sampling',
+            progress_percent=100.0,
+            log_posterior=-3.0,
+            reduced_chi2=1.0,
+            elapsed_time=5.0,
+            force_report=True,
+        )
+    )
+
+    tracker.start_sampler_post_processing()
+    tracker.stop_timer()
+    tracker.finish_tracking()
+
+    assert tracker._df_rows[-1][0] == ''
+    assert tracker._df_rows[-1][1] == ''
+    assert tracker._df_rows[-1][3] == ''
+    assert tracker._df_rows[-1][4] == 'post-processing'
diff --git a/tests/unit/easydiffraction/analysis/minimizers/test_base.py b/tests/unit/easydiffraction/analysis/minimizers/test_base.py
index 3b18f5354..4ee99f1d1 100644
--- a/tests/unit/easydiffraction/analysis/minimizers/test_base.py
+++ b/tests/unit/easydiffraction/analysis/minimizers/test_base.py
@@ -87,7 +87,8 @@ def _compute_residuals(
     assert minim.synced is True
     assert isinstance(result.parameters, list)
     assert result.parameters[0].value == 42
-    # Fitting time should be a positive float
+    assert result.fitting_time is not None
+    assert result.fitting_time >= 0.0
     assert minim.tracker.fitting_time is not None
     assert minim.tracker.fitting_time >= 0.0
 
@@ -122,6 +123,37 @@ def _compute_residuals(
     assert np.allclose(out, np.array([1.0, 2.0, 3.0]))
 
 
+def test_max_iterations_property_updates_internal_value():
+    from easydiffraction.analysis.minimizers.base import MinimizerBase
+
+    class M(MinimizerBase):
+        def __init__(self):
+            super().__init__(name='dummy', method='m', max_iterations=5)
+
+        def _prepare_solver_args(self, parameters):
+            del parameters
+            return {}
+
+        def _run_solver(self, objective_function, **kwargs):
+            del objective_function, kwargs
+            return
+
+        def _sync_result_to_parameters(self, parameters, raw_result):
+            del parameters, raw_result
+
+        def _check_success(self, raw_result):
+            del raw_result
+            return True
+
+    minimizer = M()
+
+    assert minimizer.max_iterations == 5
+
+    minimizer.max_iterations = 200
+
+    assert minimizer.max_iterations == 200
+
+
 def test_minimizer_base_fit_stops_tracking_when_solver_prep_fails():
     from easydiffraction.analysis.minimizers.base import MinimizerBase
 
@@ -201,3 +233,59 @@ def _check_success(self, raw_result):
 
     with pytest.raises(ValueError, match='prep failed'):
         minimizer.fit(parameters=[_DummyParam(1.0)], objective_function=lambda _: np.array([0.0]))
+
+
+def test_minimizer_base_stop_tracking_backfills_result_fitting_time():
+    from easydiffraction.analysis.minimizers.base import MinimizerBase
+
+    class DummyResult:
+        success = True
+
+    class DummyMinimizer(MinimizerBase):
+        def __init__(self):
+            super().__init__(name='dummy', method='m', max_iterations=5)
+
+        def _prepare_solver_args(self, parameters):
+            del parameters
+            return {'engine_parameters': {'ok': True}}
+
+        def _run_solver(self, objective_function, **kwargs):
+            residuals = objective_function(kwargs.get('engine_parameters'))
+            self.tracker.track(residuals=np.array(residuals), parameters=[1])
+            return DummyResult()
+
+        def _sync_result_to_parameters(self, parameters, raw_result):
+            del parameters, raw_result
+
+        def _check_success(self, raw_result):
+            del raw_result
+            return True
+
+        def _compute_residuals(
+            self, engine_params, parameters, structures, experiments, calculator
+        ):
+            del parameters, structures, experiments, calculator
+            assert engine_params == {'ok': True}
+            return np.array([0.0])
+
+    minimizer = DummyMinimizer()
+    params = [_DummyParam(1.0)]
+    objective = minimizer._create_objective_function(
+        parameters=params,
+        structures=None,
+        experiments=None,
+        calculator=None,
+    )
+
+    result = minimizer.fit(
+        parameters=params,
+        objective_function=objective,
+        finalize_tracking=False,
+    )
+
+    assert result.fitting_time is None
+
+    minimizer._stop_tracking()
+
+    assert result.fitting_time is not None
+    assert result.fitting_time >= 0.0
diff --git a/tests/unit/easydiffraction/analysis/minimizers/test_bumps.py b/tests/unit/easydiffraction/analysis/minimizers/test_bumps.py
index 1a73fdbcd..830a4f64f 100644
--- a/tests/unit/easydiffraction/analysis/minimizers/test_bumps.py
+++ b/tests/unit/easydiffraction/analysis/minimizers/test_bumps.py
@@ -189,6 +189,80 @@ def test_fitness_numpoints_after_nllf():
     assert fitness.numpoints() == 3
 
 
+def test_fitness_evaluation_count_can_be_reset_and_stopped():
+    from bumps.parameter import Parameter as BumpsParameter
+
+    from easydiffraction.analysis.minimizers.bumps import _EasyDiffractionFitness
+
+    bp = BumpsParameter(value=0.0, name='a')
+    fitness = _EasyDiffractionFitness([bp], lambda values: np.array([values[0]]))
+
+    fitness.residuals()
+    fitness.residuals()
+    assert fitness.evaluation_count == 2
+
+    fitness.reset_evaluation_count()
+    assert fitness.evaluation_count == 0
+
+    fitness.stop_counting_evaluations()
+    fitness.residuals()
+    assert fitness.evaluation_count == 0
+
+
+def test_fitness_raises_when_max_evaluations_is_reached():
+    from bumps.parameter import Parameter as BumpsParameter
+
+    from easydiffraction.analysis.minimizers.bumps import _BumpsEvaluationLimitError
+    from easydiffraction.analysis.minimizers.bumps import _EasyDiffractionFitness
+
+    bp = BumpsParameter(value=2.0, name='a')
+    fitness = _EasyDiffractionFitness(
+        [bp],
+        lambda values: np.array([values[0]]),
+        max_evaluations=2,
+    )
+
+    fitness.residuals()
+    fitness.residuals()
+
+    with pytest.raises(_BumpsEvaluationLimitError) as exc_info:
+        fitness.residuals()
+
+    assert exc_info.value.evaluation_count == 2
+    np.testing.assert_array_equal(exc_info.value.parameter_values, np.array([2.0]))
+    np.testing.assert_array_equal(exc_info.value.residuals, np.array([2.0]))
+
+
+def test_bumps_progress_monitor_reports_evaluation_count():
+    from easydiffraction.analysis.minimizers.bumps import _EasyDiffractionFitness
+    from easydiffraction.analysis.minimizers.bumps import _BumpsProgressMonitor
+
+    tracker = MagicMock()
+    fitness = _EasyDiffractionFitness([], lambda values: np.array([]))
+    fitness.reset_evaluation_count()
+    fitness._evaluation_count = 63
+    monitor = _BumpsProgressMonitor(
+        tracker=tracker,
+        fitness=fitness,
+        n_points=20,
+        n_parameters=4,
+    )
+
+    monitor(
+        types.SimpleNamespace(
+            step=[7],
+            value=[8.0],
+            time=[1.5],
+        )
+    )
+
+    tracker.track_fit_progress.assert_called_once_with(
+        iteration=63,
+        reduced_chi2=pytest.approx(1.0),
+        elapsed_time=1.5,
+    )
+
+
 def test_fitness_update_is_noop():
     from easydiffraction.analysis.minimizers.bumps import _EasyDiffractionFitness
 
@@ -232,6 +306,8 @@ def test_run_solver_returns_optimize_result():
             bumps_params=[bp1, bp2],
         )
 
+    assert len(mock_driver_cls.call_args.kwargs['monitors']) == 1
+
     assert isinstance(res, OptimizeResult)
     assert res.success is True
     np.testing.assert_array_almost_equal(res.x, [1.5, 2.5])
@@ -265,6 +341,51 @@ def test_run_solver_failure():
     assert res.status == -1
 
 
+def test_run_solver_stops_at_max_evaluations():
+    from easydiffraction.analysis.minimizers.bumps import BumpsMinimizer
+    from easydiffraction.analysis.minimizers.bumps import _BumpsEvaluationLimitError
+
+    m = BumpsMinimizer(max_iterations=50)
+    m.tracker = MagicMock()
+    m.tracker._current_elapsed_time.return_value = 1.25
+
+    fake_fitter = types.SimpleNamespace(id='lm')
+    limit_error = _BumpsEvaluationLimitError(
+        evaluation_count=50,
+        parameter_values=np.array([1.5]),
+        residuals=np.array([2.0, 2.0]),
+    )
+
+    with (
+        patch('easydiffraction.analysis.minimizers.bumps.FitDriver') as mock_driver_cls,
+        patch('easydiffraction.analysis.minimizers.bumps.FitProblem'),
+        patch('easydiffraction.analysis.minimizers.bumps.FITTERS', [fake_fitter]),
+        patch.object(m, '_compute_covariance', return_value=(None, None)) as mock_covariance,
+    ):
+        driver_instance = mock_driver_cls.return_value
+        driver_instance.fit.side_effect = limit_error
+        driver_instance.clip = MagicMock()
+
+        from bumps.parameter import Parameter as BumpsParameter
+
+        bp = BumpsParameter(value=1.5, name='a')
+        result = m._run_solver(
+            lambda values: np.array([values[0], values[0]]),
+            bumps_params=[bp],
+        )
+
+    assert result.success is False
+    assert result.status == 5
+    assert result.message == 'maximum number of residual evaluations reached'
+    np.testing.assert_array_equal(result.x, np.array([1.5]))
+    mock_covariance.assert_not_called()
+    m.tracker.track_fit_progress.assert_called_once_with(
+        iteration=50,
+        reduced_chi2=pytest.approx(8.0),
+        elapsed_time=1.25,
+    )
+
+
 # -- _sync_result_to_parameters tests -----------------------------------------
 
 
diff --git a/tests/unit/easydiffraction/analysis/minimizers/test_bumps_dream.py b/tests/unit/easydiffraction/analysis/minimizers/test_bumps_dream.py
index d17dce4fa..3a23f5577 100644
--- a/tests/unit/easydiffraction/analysis/minimizers/test_bumps_dream.py
+++ b/tests/unit/easydiffraction/analysis/minimizers/test_bumps_dream.py
@@ -53,6 +53,28 @@ def test_type_info_and_default_init():
     assert minimizer.steps == 3000
 
 
+def test_dream_uses_steps_instead_of_max_iterations():
+    from easydiffraction.analysis.minimizers.bumps_dream import BumpsDreamMinimizer
+
+    minimizer = BumpsDreamMinimizer()
+
+    with pytest.raises(
+        AttributeError,
+        match=r"DREAM sampler uses 'steps' instead of 'max_iterations'\.",
+    ):
+        _ = minimizer.max_iterations
+
+    with pytest.raises(
+        AttributeError,
+        match=r"DREAM sampler uses 'steps' instead of 'max_iterations'\.",
+    ):
+        minimizer.max_iterations = 300
+
+    minimizer.steps = 300
+
+    assert minimizer.steps == 300
+
+
 def test_dream_progress_monitor_allocates_rows_by_phase_ratio():
     from easydiffraction.analysis.minimizers.bumps_dream import _DreamProgressMonitor
 
diff --git a/tests/unit/easydiffraction/analysis/minimizers/test_dfols.py b/tests/unit/easydiffraction/analysis/minimizers/test_dfols.py
index 7d94bd04e..68437dd2a 100644
--- a/tests/unit/easydiffraction/analysis/minimizers/test_dfols.py
+++ b/tests/unit/easydiffraction/analysis/minimizers/test_dfols.py
@@ -43,10 +43,12 @@ def __init__(self):
 
     def fake_solve(fun, x0, bounds, maxfun):
         # Verify we pass reasonable arguments
+        del fun
         assert isinstance(x0, np.ndarray)
         assert x0.shape[0] == 2
         assert isinstance(bounds, tuple)
         assert all(isinstance(b, np.ndarray) for b in bounds)
+        assert maxfun == 10
         return FakeRes()
 
     monkeypatch.setattr(mod, 'solve', fake_solve)
@@ -63,3 +65,5 @@ def fake_solve(fun, x0, bounds, maxfun):
     assert params[0].uncertainty is None
     assert params[1].uncertainty is None
     assert minim._check_success(res) is True
+    assert minim.max_iterations == 10
+    assert not hasattr(minim, 'steps')
diff --git a/tests/unit/easydiffraction/analysis/minimizers/test_lmfit.py b/tests/unit/easydiffraction/analysis/minimizers/test_lmfit.py
index f41e98965..7091fdcd4 100644
--- a/tests/unit/easydiffraction/analysis/minimizers/test_lmfit.py
+++ b/tests/unit/easydiffraction/analysis/minimizers/test_lmfit.py
@@ -77,3 +77,38 @@ def __init__(self):
     assert params[1].value == 20.0
     assert params[1].uncertainty == 1.0
     assert minim._check_success(res) is True
+
+
+def test_lmfit_max_iterations_is_user_facing_iteration_setting(monkeypatch):
+    from easydiffraction.analysis.minimizers.lmfit import LmfitMinimizer
+
+    import easydiffraction.analysis.minimizers.lmfit as lm
+
+    observed_max_nfev = {}
+
+    def fake_minimize(
+        objective_function,
+        *,
+        params,
+        method,
+        nan_policy,
+        max_nfev,
+    ):
+        del objective_function, params, method, nan_policy
+        observed_max_nfev['value'] = max_nfev
+        return types.SimpleNamespace(success=True, params={})
+
+    monkeypatch.setattr(
+        lm,
+        'lmfit',
+        types.SimpleNamespace(Parameters=lm.lmfit.Parameters, minimize=fake_minimize),
+    )
+
+    minimizer = LmfitMinimizer()
+
+    minimizer.max_iterations = 300
+    minimizer._run_solver(lambda *args, **kwargs: np.array([0.0]), engine_parameters=object())
+
+    assert minimizer.max_iterations == 300
+    assert observed_max_nfev['value'] == 300
+    assert not hasattr(minimizer, 'steps')
diff --git a/tests/unit/easydiffraction/analysis/test_analysis_coverage.py b/tests/unit/easydiffraction/analysis/test_analysis_coverage.py
index 13072544a..8a741fef7 100644
--- a/tests/unit/easydiffraction/analysis/test_analysis_coverage.py
+++ b/tests/unit/easydiffraction/analysis/test_analysis_coverage.py
@@ -2,6 +2,10 @@
 # SPDX-License-Identifier: BSD-3-Clause
 """Additional unit tests for analysis.py to cover patch gaps."""
 
+from types import SimpleNamespace
+
+import numpy as np
+
 
 def _make_project():
     class ExpCol:
@@ -199,3 +203,104 @@ class FakeResults:
         assert 'expt1' in a._parameter_snapshots
         assert a._parameter_snapshots['expt1']['p1']['value'] == 1.23
         assert a._parameter_snapshots['expt1']['p1']['uncertainty'] == 0.01
+
+
+class TestBayesianProjection:
+    def test_single_parameter_projection_persists_distribution_and_predictive_caches(self):
+        from easydiffraction.analysis.analysis import Analysis
+        from easydiffraction.analysis.fit_helpers.bayesian import BayesianFitResults
+        from easydiffraction.analysis.fit_helpers.bayesian import PosteriorParameterSummary
+        from easydiffraction.analysis.fit_helpers.bayesian import PosteriorPredictiveSummary
+        from easydiffraction.analysis.fit_helpers.bayesian import PosteriorSamples
+
+        class Plotter:
+            @staticmethod
+            def _posterior_parameter_bounds(*, fit_results, parameter_name):
+                del fit_results, parameter_name
+                return 0.5, 1.5
+
+            @staticmethod
+            def _posterior_density_curve(values, *, lower_bound, upper_bound):
+                del values
+                return (
+                    np.asarray([lower_bound, upper_bound], dtype=float),
+                    np.asarray([0.25, 0.75], dtype=float),
+                )
+
+            @staticmethod
+            def _resolve_x_axis(experiment_type, _axis_name):
+                del experiment_type
+                return np.asarray([1.0, 2.0], dtype=float), 'two_theta', None, None, None
+
+            @staticmethod
+            def _build_posterior_predictive_summary(
+                *,
+                fit_results,
+                experiment,
+                expt_name,
+                x_axis,
+                include_draws,
+            ):
+                del fit_results, experiment, x_axis, include_draws
+                return PosteriorPredictiveSummary(
+                    experiment_name=expt_name,
+                    x_axis_name='two_theta',
+                    x=np.asarray([1.0, 2.0], dtype=float),
+                    best_sample_prediction=np.asarray([3.0, 4.0], dtype=float),
+                    lower_95=np.asarray([2.5, 3.5], dtype=float),
+                    upper_95=np.asarray([3.5, 4.5], dtype=float),
+                )
+
+        class Experiments:
+            names = ['hrpt']
+
+            def __getitem__(self, name):
+                del name
+                return SimpleNamespace(type='powder')
+
+        project = SimpleNamespace(
+            experiments=Experiments(),
+            structures=object(),
+            rendering=SimpleNamespace(plotter=Plotter()),
+            _varname='proj',
+        )
+        analysis = Analysis(project=project)
+
+        results = BayesianFitResults(
+            success=True,
+            parameters=[],
+            posterior_samples=PosteriorSamples(
+                parameter_names=['alpha'],
+                parameter_samples=np.asarray([[[1.0]], [[1.2]]], dtype=float),
+            ),
+            posterior_parameter_summaries=[
+                PosteriorParameterSummary(
+                    unique_name='alpha',
+                    display_name='Alpha',
+                    best_sample_value=1.2,
+                    median=1.1,
+                    standard_deviation=0.1,
+                    interval_68=(1.0, 1.2),
+                    interval_95=(0.9, 1.3),
+                )
+            ],
+            posterior_predictive={},
+            sampler_settings={},
+            convergence_diagnostics={},
+        )
+
+        analysis._store_bayesian_result_projection(results)
+
+        assert analysis.bayesian_result.has_distribution_cache.value is True
+        assert analysis.bayesian_result.has_pair_cache.value is False
+        assert analysis.bayesian_result.has_posterior_predictive.value is True
+        assert np.allclose(
+            analysis._persisted_fit_state_sidecar['distribution_caches']['alpha']['x'],
+            np.asarray([0.5, 1.5], dtype=float),
+        )
+        assert np.allclose(
+            analysis._persisted_fit_state_sidecar['predictive_datasets']['hrpt'][
+                'best_sample_prediction'
+            ],
+            np.asarray([3.0, 4.0], dtype=float),
+        )
diff --git a/tests/unit/easydiffraction/analysis/test_enums.py b/tests/unit/easydiffraction/analysis/test_enums.py
index fe854e536..75e1ca3ec 100644
--- a/tests/unit/easydiffraction/analysis/test_enums.py
+++ b/tests/unit/easydiffraction/analysis/test_enums.py
@@ -2,6 +2,8 @@
 # SPDX-License-Identifier: BSD-3-Clause
 """Tests for analysis/enums.py."""
 
+from easydiffraction.analysis.enums import FitCorrelationSourceEnum
+from easydiffraction.analysis.enums import FitResultKindEnum
 from easydiffraction.analysis.enums import FitModeEnum
 
 
@@ -20,3 +22,15 @@ def test_fit_mode_enum_descriptions():
         description = member.description()
         assert isinstance(description, str)
         assert description
+
+
+def test_fit_result_kind_enum_members_and_default():
+    assert FitResultKindEnum.DETERMINISTIC == 'deterministic'
+    assert FitResultKindEnum.BAYESIAN == 'bayesian'
+    assert FitResultKindEnum.default() is FitResultKindEnum.DETERMINISTIC
+
+
+def test_fit_correlation_source_enum_members_and_default():
+    assert FitCorrelationSourceEnum.DETERMINISTIC == 'deterministic'
+    assert FitCorrelationSourceEnum.POSTERIOR == 'posterior'
+    assert FitCorrelationSourceEnum.default() is FitCorrelationSourceEnum.DETERMINISTIC
diff --git a/tests/unit/easydiffraction/analysis/test_fitting.py b/tests/unit/easydiffraction/analysis/test_fitting.py
index 0f7ef88b1..83272f9fa 100644
--- a/tests/unit/easydiffraction/analysis/test_fitting.py
+++ b/tests/unit/easydiffraction/analysis/test_fitting.py
@@ -1,6 +1,10 @@
 # SPDX-FileCopyrightText: 2025 EasyScience contributors <https://github.com/easyscience>
 # SPDX-License-Identifier: BSD-3-Clause
 
+from types import SimpleNamespace
+
+from easydiffraction.utils.enums import VerbosityEnum
+
 
 def test_module_import():
     import easydiffraction.analysis.fitting as MUT
@@ -71,7 +75,11 @@ class DummyExperiment:
         parameters = []
 
     class MockFitResults:
-        pass
+        def __init__(self):
+            self.message = ''
+            self.iterations = 0
+            self.chi_square = None
+            self.engine_result = object()
 
     class DummyMin:
         tracker = type('T', (), {'track': staticmethod(lambda a, b: a)})()
@@ -82,6 +90,9 @@ def fit(self, params, obj, verbosity=None, **kwargs):
         def _sync_result_to_parameters(self, params, engine_params):
             pass
 
+        def _stop_tracking(self):
+            return None
+
     f = Fitter()
     f.minimizer = DummyMin()
 
@@ -101,3 +112,174 @@ def mock_process(*args, **kwargs):
         'Use Analysis.show_fit_results() instead.'
     )
     assert f.results is not None, 'Fitter.fit() should still set results'
+
+
+def test_fitter_fit_defers_minimizer_tracking_until_postprocessing(monkeypatch):
+    from easydiffraction.analysis.fit_helpers.bayesian import BayesianFitResults
+    from easydiffraction.analysis.fitting import Fitter
+
+    class DummyParam:
+        value = 1.0
+        uncertainty = 0.1
+        _fit_start_value = None
+
+    class DummyStructure:
+        _need_categories_update = False
+
+        def _update_categories(self):
+            return None
+
+    class DummyStructures:
+        def __iter__(self):
+            return iter([DummyStructure()])
+
+    class DummyExperiment:
+        parameters = []
+
+    class DummyMin:
+        def __init__(self):
+            self.fit_calls: list[dict[str, object]] = []
+            self.stop_calls = 0
+            self.tracker = SimpleNamespace(track=lambda residuals, parameters: residuals)
+
+        def fit(self, params, obj, verbosity=None, **kwargs):
+            del params, obj
+            self.fit_calls.append({'verbosity': verbosity, **kwargs})
+            return BayesianFitResults(
+                success=True,
+                reduced_chi_square=1.2,
+                convergence_diagnostics={'converged': False},
+                sampler_settings={'steps': 300},
+                best_log_posterior=-10.0,
+            )
+
+        def _stop_tracking(self):
+            self.stop_calls += 1
+
+    analysis_events: list[str] = []
+    analysis = SimpleNamespace(
+        _capture_fit_parameter_state=lambda params: analysis_events.append('capture'),
+        _store_fit_result_projection=lambda results, experiments, fitted_parameters: (
+            analysis_events.append('store')
+        ),
+    )
+
+    fitter = Fitter()
+    fitter.minimizer = DummyMin()
+    monkeypatch.setattr(
+        fitter,
+        '_collect_fit_parameters',
+        lambda structures, experiments: [DummyParam()],
+    )
+
+    fitter.fit(
+        structures=DummyStructures(),
+        experiments=[DummyExperiment()],
+        analysis=analysis,
+        verbosity=VerbosityEnum.FULL,
+    )
+
+    assert fitter.minimizer.fit_calls[0]['finalize_tracking'] is False
+    assert fitter.minimizer.stop_calls == 1
+    assert analysis_events == ['capture', 'store']
+
+
+def test_fitter_fit_stops_tracking_when_minimizer_fit_raises(monkeypatch):
+    import pytest
+
+    from easydiffraction.analysis.fitting import Fitter
+
+    class DummyParam:
+        value = 1.0
+        _fit_start_value = None
+
+    class DummyStructure:
+        _need_categories_update = False
+
+        def _update_categories(self):
+            return None
+
+    class DummyStructures:
+        def __iter__(self):
+            return iter([DummyStructure()])
+
+    class DummyExperiment:
+        parameters = []
+
+    class DummyMin:
+        def __init__(self):
+            self.stop_calls = 0
+            self.tracker = SimpleNamespace(track=lambda residuals, parameters: residuals)
+
+        def fit(self, params, obj, verbosity=None, **kwargs):
+            del params, obj, verbosity, kwargs
+            msg = 'fit failed'
+            raise RuntimeError(msg)
+
+        def _stop_tracking(self):
+            self.stop_calls += 1
+
+    fitter = Fitter()
+    fitter.minimizer = DummyMin()
+    monkeypatch.setattr(
+        fitter,
+        '_collect_fit_parameters',
+        lambda structures, experiments: [DummyParam()],
+    )
+
+    with pytest.raises(RuntimeError, match='fit failed'):
+        fitter.fit(
+            structures=DummyStructures(),
+            experiments=[DummyExperiment()],
+            verbosity=VerbosityEnum.FULL,
+        )
+
+    assert fitter.minimizer.stop_calls == 1
+
+
+def test_residual_function_skips_tracker_for_solver_monitored_minimizer(monkeypatch):
+    import numpy as np
+
+    from easydiffraction.analysis.fitting import Fitter
+
+    class DummyExperiment:
+        def _update_categories(self, *, called_by_minimizer=False):
+            del called_by_minimizer
+            return
+
+    class DummyMin:
+        def __init__(self):
+            self.tracker = SimpleNamespace(
+                track=lambda residuals, parameters: (_ for _ in ()).throw(
+                    AssertionError('tracker.track should not be called')
+                )
+            )
+
+        def _sync_result_to_parameters(self, parameters, engine_params):
+            del parameters, engine_params
+
+        def _tracks_progress_via_solver_monitor(self):
+            return True
+
+    fitter = Fitter()
+    fitter.minimizer = DummyMin()
+
+    monkeypatch.setattr(
+        'easydiffraction.analysis.fitting.intensity_category_for',
+        lambda experiment: SimpleNamespace(
+            intensity_calc=np.array([1.0]),
+            intensity_meas=np.array([2.0]),
+            intensity_meas_su=np.array([1.0]),
+        ),
+    )
+
+    residuals = fitter._residual_function(
+        engine_params={},
+        parameters=[],
+        structures=[],
+        experiments=[DummyExperiment()],
+        weights=None,
+        analysis=None,
+    )
+
+    np.testing.assert_allclose(residuals, np.array([1.0]))
diff --git a/tests/unit/easydiffraction/analysis/test_sequential.py b/tests/unit/easydiffraction/analysis/test_sequential.py
index ab2327c94..2be4c19f3 100644
--- a/tests/unit/easydiffraction/analysis/test_sequential.py
+++ b/tests/unit/easydiffraction/analysis/test_sequential.py
@@ -137,7 +137,7 @@ def _run_non_silent_fit(monkeypatch, tmp_path, *, verbosity, is_jupyter):
     del is_jupyter  # legacy parameter, no longer affects behavior
 
     analysis = SimpleNamespace(
-        project=SimpleNamespace(verbosity=verbosity),
+        project=SimpleNamespace(verbosity=SimpleNamespace(fit=SimpleNamespace(value=verbosity))),
         fitter=SimpleNamespace(selection='lmfit'),
     )
 
@@ -198,7 +198,7 @@ def test_full_header_order(self):
 class TestCsvWriteAndAppend:
     def test_write_creates_file_with_header(self, tmp_path):
         csv_path = tmp_path / 'results.csv'
-        header = ['file_path', 'chi_squared', 'param_a']
+        header = ['file_path', 'fit_result.reduced_chi_square', 'param_a']
         _write_csv_header(csv_path, header)
 
         with csv_path.open() as f:
@@ -336,19 +336,17 @@ def test_returns_fitted_file_paths(self, tmp_path):
             [
                 {
                     'file_path': str(project_dir / 'experiments' / 'a.dat'),
-                    'fit_success': 'True',
-                    'chi_squared': '5.0',
-                    'reduced_chi_squared': '2.5',
-                    'n_iterations': '10',
+                    'fit_result.success': 'True',
+                    'fit_result.reduced_chi_square': '2.5',
+                    'fit_result.iterations': '10',
                     'cell.a': '3.89',
                     'cell.a.uncertainty': '0.01',
                 },
                 {
                     'file_path': str(project_dir / 'experiments' / 'b.dat'),
-                    'fit_success': 'False',
-                    'chi_squared': '',
-                    'reduced_chi_squared': '',
-                    'n_iterations': '0',
+                    'fit_result.success': 'False',
+                    'fit_result.reduced_chi_square': '',
+                    'fit_result.iterations': '0',
                     'cell.a': '',
                     'cell.a.uncertainty': '',
                 },
@@ -374,10 +372,9 @@ def test_resolves_legacy_repo_relative_paths(self, tmp_path, monkeypatch):
             writer.writeheader()
             writer.writerow({
                 'file_path': 'projects/cosio/experiments/d20_scan/scan_001.dat',
-                'fit_success': 'True',
-                'chi_squared': '5.0',
-                'reduced_chi_squared': '2.5',
-                'n_iterations': '10',
+                'fit_result.success': 'True',
+                'fit_result.reduced_chi_square': '2.5',
+                'fit_result.iterations': '10',
                 'cell.a': '3.89',
                 'cell.a.uncertainty': '0.01',
             })
@@ -397,19 +394,17 @@ def test_returns_last_successful_params(self, tmp_path):
             [
                 {
                     'file_path': 'a.dat',
-                    'fit_success': 'True',
-                    'chi_squared': '5.0',
-                    'reduced_chi_squared': '2.5',
-                    'n_iterations': '10',
+                    'fit_result.success': 'True',
+                    'fit_result.reduced_chi_square': '2.5',
+                    'fit_result.iterations': '10',
                     'cell.a': '3.89',
                     'cell.a.uncertainty': '0.01',
                 },
                 {
                     'file_path': 'b.dat',
-                    'fit_success': 'True',
-                    'chi_squared': '4.0',
-                    'reduced_chi_squared': '2.0',
-                    'n_iterations': '8',
+                    'fit_result.success': 'True',
+                    'fit_result.reduced_chi_square': '2.0',
+                    'fit_result.iterations': '8',
                     'cell.a': '3.90',
                     'cell.a.uncertainty': '0.02',
                 },
@@ -436,10 +431,9 @@ def test_skips_meta_columns_and_diffrn_and_uncertainty(self, tmp_path):
             [
                 {
                     'file_path': 'a.dat',
-                    'fit_success': 'True',
-                    'chi_squared': '5.0',
-                    'reduced_chi_squared': '2.5',
-                    'n_iterations': '10',
+                    'fit_result.success': 'True',
+                    'fit_result.reduced_chi_square': '2.5',
+                    'fit_result.iterations': '10',
                     'diffrn.temp': '300',
                     'cell.a': '3.89',
                     'cell.a.uncertainty': '0.01',
@@ -452,7 +446,7 @@ def test_skips_meta_columns_and_diffrn_and_uncertainty(self, tmp_path):
         assert 'cell.a' in params
         # Meta columns, diffrn, and uncertainty should be excluded
         assert 'file_path' not in params
-        assert 'fit_success' not in params
+        assert 'fit_result.success' not in params
         assert 'diffrn.temp' not in params
         assert 'cell.a.uncertainty' not in params
 
@@ -466,10 +460,9 @@ def test_returns_none_params_when_no_successful_rows(self, tmp_path):
             [
                 {
                     'file_path': 'a.dat',
-                    'fit_success': 'False',
-                    'chi_squared': '',
-                    'reduced_chi_squared': '',
-                    'n_iterations': '0',
+                    'fit_result.success': 'False',
+                    'fit_result.reduced_chi_square': '',
+                    'fit_result.iterations': '0',
                     'cell.a': '',
                     'cell.a.uncertainty': '',
                 },
@@ -537,15 +530,15 @@ def update(self, *, label=None, content=None):
         [
             {
                 'file_path': _TEST_SCAN_001,
-                'fit_success': True,
-                'reduced_chi_squared': 4.0,
-                'n_iterations': 11,
+                'fit_result.success': True,
+                'fit_result.reduced_chi_square': 4.0,
+                'fit_result.iterations': 11,
             },
             {
                 'file_path': _TEST_SCAN_002,
-                'fit_success': False,
-                'reduced_chi_squared': None,
-                'n_iterations': 0,
+                'fit_result.success': False,
+                'fit_result.reduced_chi_square': None,
+                'fit_result.iterations': 0,
             },
         ],
         progress,
@@ -608,7 +601,7 @@ def test_fit_sequential_non_silent_starts_indicator_with_progress_table(
     assert events[8] == ('stop',)
     assert events[9:] == [
         ('console_print', ('✅ Sequential fitting complete: 1 files processed.',), {}),
-        ('console_print', (f'📄 Results saved to: {tmp_path / "results.csv"}',), {}),
+        ('console_print', (f'📄 Results saved to:\n{tmp_path / "results.csv"}',), {}),
     ]
 
 
@@ -631,9 +624,9 @@ def map(self, func, templates, paths):
             for path in paths:
                 yield {
                     'file_path': path,
-                    'fit_success': True,
-                    'reduced_chi_squared': 1.0,
-                    'n_iterations': 5,
+                    'fit_result.success': True,
+                    'fit_result.reduced_chi_square': 1.0,
+                    'fit_result.iterations': 5,
                     'params': {'cell.a': 4.0},
                 }
 
@@ -728,7 +721,7 @@ def fake_run_fit_loop(
     monkeypatch.setattr(sequential_mod, '_restore_main_state', lambda *args: None)
 
     analysis = SimpleNamespace(
-        project=SimpleNamespace(verbosity='silent'),
+        project=SimpleNamespace(verbosity=SimpleNamespace(fit=SimpleNamespace(value='silent'))),
         fitter=SimpleNamespace(selection='lmfit'),
     )
 
diff --git a/tests/unit/easydiffraction/display/test_plotting.py b/tests/unit/easydiffraction/display/test_plotting.py
index 461e85ce2..d5b6ae76c 100644
--- a/tests/unit/easydiffraction/display/test_plotting.py
+++ b/tests/unit/easydiffraction/display/test_plotting.py
@@ -1,6 +1,7 @@
 # SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
 # SPDX-License-Identifier: BSD-3-Clause
 
+import csv
 import re
 from types import MethodType
 from types import SimpleNamespace
@@ -66,6 +67,78 @@ def test_plotter_factory_supported_and_unsupported():
         PlotterFactory.create('nope')
 
 
+@pytest.mark.parametrize(
+    ('descriptor_name', 'column_values', 'expected_y'),
+    [
+        ('reduced_chi_square', ['1.5', '2.5'], [1.5, 2.5]),
+        ('iterations', ['5', '8'], [5.0, 8.0]),
+        ('success', ['True', 'False'], [1.0, 0.0]),
+    ],
+)
+def test_plot_param_series_reads_fit_result_columns_from_csv(
+    monkeypatch,
+    tmp_path,
+    descriptor_name,
+    column_values,
+    expected_y,
+):
+    from easydiffraction.display.plotting import Plotter
+    from easydiffraction.project.project import Project
+
+    project = Project(name='series')
+    project.info.path = tmp_path
+
+    analysis_dir = tmp_path / 'analysis'
+    analysis_dir.mkdir(parents=True)
+    csv_path = analysis_dir / 'results.csv'
+    with csv_path.open('w', newline='', encoding='utf-8') as handle:
+        writer = csv.DictWriter(
+            handle,
+            fieldnames=[
+                'file_path',
+                'fit_result.reduced_chi_square',
+                'fit_result.success',
+                'fit_result.iterations',
+            ],
+        )
+        writer.writeheader()
+        writer.writerow({
+            'file_path': 'a.dat',
+            'fit_result.reduced_chi_square': column_values[0],
+            'fit_result.success': column_values[0] if descriptor_name == 'success' else 'True',
+            'fit_result.iterations': column_values[0] if descriptor_name == 'iterations' else '5',
+        })
+        writer.writerow({
+            'file_path': 'b.dat',
+            'fit_result.reduced_chi_square': column_values[1],
+            'fit_result.success': column_values[1] if descriptor_name == 'success' else 'False',
+            'fit_result.iterations': column_values[1] if descriptor_name == 'iterations' else '8',
+        })
+
+    captured: dict[str, object] = {}
+
+    plotter = Plotter()
+    plotter._set_project(project)
+
+    def fake_plot_scatter(*, x, y, sy, axes_labels, title, height):
+        captured['x'] = x
+        captured['y'] = y
+        captured['sy'] = sy
+        captured['axes_labels'] = axes_labels
+        captured['title'] = title
+        captured['height'] = height
+
+    monkeypatch.setattr(plotter._backend, 'plot_scatter', fake_plot_scatter)
+
+    plotter.plot_param_series(getattr(project.analysis.fit_result, descriptor_name))
+
+    assert captured['x'] == [1, 2]
+    assert captured['y'] == expected_y
+    assert captured['sy'] == [0.0, 0.0]
+    assert captured['axes_labels'] == ['Experiment No.', 'Parameter value']
+    assert captured['title'] == f"Parameter 'fit_result.{descriptor_name}' across fit results"
+
+
 def test_plotter_error_paths_and_filtering(capsys, monkeypatch):
     from easydiffraction.datablocks.experiment.item.enums import BeamModeEnum
     from easydiffraction.datablocks.experiment.item.enums import SampleFormEnum
diff --git a/tests/unit/easydiffraction/io/cif/test_serialize.py b/tests/unit/easydiffraction/io/cif/test_serialize.py
index 3a24b6ce2..375e5aa86 100644
--- a/tests/unit/easydiffraction/io/cif/test_serialize.py
+++ b/tests/unit/easydiffraction/io/cif/test_serialize.py
@@ -27,7 +27,7 @@ def __init__(self):
             self.value = 3
 
     p = P()
-    assert MUT.param_to_cif(p) == '_x.y 3.'
+    assert MUT.param_to_cif(p) == '_x.y 3'
 
 
 def test_format_param_value_with_uncertainty_uses_two_sig_digits():
diff --git a/tests/unit/easydiffraction/io/cif/test_serialize_more.py b/tests/unit/easydiffraction/io/cif/test_serialize_more.py
index f526f21a8..3c9db1bc1 100644
--- a/tests/unit/easydiffraction/io/cif/test_serialize_more.py
+++ b/tests/unit/easydiffraction/io/cif/test_serialize_more.py
@@ -34,7 +34,7 @@ def __init__(self):
 
     out = MUT.datablock_item_to_cif(DB())
     assert out.startswith('data_block1')
-    assert '_aa 42.' in out
+    assert '_aa 42' in out
     assert 'loop_' in out
     assert '_aa' in out
     assert '7' in out
@@ -80,7 +80,7 @@ def __init__(self):
             self.empty_coll = CategoryCollection(item_type=Item)
 
     out = MUT.datablock_item_to_cif(DB())
-    assert out == 'data_block1\n\n_aa 42.\n\nloop_\n_aa\n7.'
+    assert out == 'data_block1\n\n_aa 42\n\nloop_\n_aa\n7'
     assert '\n\n\n' not in out
 
 
@@ -107,11 +107,24 @@ def test_project_info_to_cif_contains_core_fields():
     info = ProjectInfo(name='p1', title='My Title', description='Some description text')
     out = MUT.project_info_to_cif(info)
     assert '_project.id               p1' in out
-    assert '_project.title' in out
-    assert 'My Title' in out
-    assert '_project.description' in out
-    assert '_project.created' in out
-    assert '_project.last_modified' in out
+    assert '_project.title            "My Title"' in out
+    assert '_project.description      "Some description text"' in out
+    assert '_project.created          "' in out
+    assert '_project.last_modified    "' in out
+
+
+def test_project_info_to_cif_wraps_long_description_as_text_field():
+    import easydiffraction.io.cif.serialize as MUT
+    from easydiffraction.project.project_info import ProjectInfo
+
+    description = ' '.join(['long'] * 20)
+    info = ProjectInfo(name='p1', title='My Title', description=description)
+
+    out = MUT.project_info_to_cif(info)
+
+    assert '_project.description      ' in out
+    assert '\n;\n' in out
+    assert 'long long long long long long long long long long long long' in out
 
 
 def test_experiment_to_cif_with_and_without_data():
@@ -159,7 +172,7 @@ def as_cif(self):
 
     out_without = MUT.experiment_to_cif(Exp(''))
     assert out_without.startswith('data_expA')
-    assert out_without.endswith('1.')
+    assert out_without.endswith('1')
 
 
 def test_analysis_to_cif_renders_all_sections():
diff --git a/tests/unit/easydiffraction/io/test_results_sidecar.py b/tests/unit/easydiffraction/io/test_results_sidecar.py
new file mode 100644
index 000000000..556e21e5a
--- /dev/null
+++ b/tests/unit/easydiffraction/io/test_results_sidecar.py
@@ -0,0 +1,138 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Tests for persisted Bayesian results sidecars."""
+
+from __future__ import annotations
+
+from pathlib import Path
+from types import SimpleNamespace
+
+import numpy as np
+
+
+def _analysis_with_predictive_sidecar() -> object:
+    from easydiffraction.analysis.categories.bayesian_predictive_datasets.default import (
+        BayesianPredictiveDatasetPaths,
+        BayesianPredictiveDatasets,
+    )
+    from easydiffraction.analysis.categories.bayesian_result.default import BayesianResult
+    from easydiffraction.analysis.categories.fit_result.default import FitResult
+
+    fit_result = FitResult()
+    fit_result._set_result_kind('bayesian')
+    bayesian_result = BayesianResult()
+    bayesian_result._set_has_posterior_predictive(value=True)
+    predictive = BayesianPredictiveDatasets()
+    predictive.create(
+        experiment_name='hrpt',
+        x_axis_name='two_theta',
+        paths=BayesianPredictiveDatasetPaths(
+            x_path='/predictive/hrpt/x',
+            best_sample_prediction_path='/predictive/hrpt/best_sample_prediction',
+            lower_95_path='/predictive/hrpt/lower_95',
+            upper_95_path='/predictive/hrpt/upper_95',
+        ),
+        n_x=2,
+        n_draws_cached=0,
+    )
+    return SimpleNamespace(
+        fit_result=fit_result,
+        bayesian_result=bayesian_result,
+        bayesian_convergence=SimpleNamespace(
+            n_draws=SimpleNamespace(value=0),
+            n_chains=SimpleNamespace(value=0),
+            n_parameters=SimpleNamespace(value=0),
+        ),
+        bayesian_distribution_caches=[],
+        bayesian_pair_caches=[],
+        bayesian_predictive_datasets=predictive,
+        fit_results=SimpleNamespace(
+            posterior_predictive={
+                'hrpt': SimpleNamespace(
+                    experiment_name='hrpt',
+                    x_axis_name='two_theta',
+                    x=np.asarray([1.0, 2.0]),
+                    best_sample_prediction=np.asarray([3.0, 4.0]),
+                    lower_95=np.asarray([2.5, 3.5]),
+                    upper_95=np.asarray([3.5, 4.5]),
+                    lower_68=None,
+                    upper_68=None,
+                    draws=None,
+                )
+            }
+        ),
+        _persisted_fit_state_sidecar={},
+        _has_persisted_fit_state=lambda: True,
+    )
+
+
+def test_write_and_read_analysis_results_sidecar_round_trip_predictive(tmp_path):
+    from easydiffraction.io.results_sidecar import read_analysis_results_sidecar
+    from easydiffraction.io.results_sidecar import write_analysis_results_sidecar
+
+    analysis_dir = Path(tmp_path)
+    analysis = _analysis_with_predictive_sidecar()
+
+    write_analysis_results_sidecar(analysis=analysis, analysis_dir=analysis_dir)
+
+    sidecar_path = analysis_dir / 'results.h5'
+    assert sidecar_path.is_file()
+
+    restored = _analysis_with_predictive_sidecar()
+    restored.fit_results = None
+    read_analysis_results_sidecar(analysis=restored, analysis_dir=analysis_dir)
+
+    assert 'predictive_datasets' in restored._persisted_fit_state_sidecar
+    dataset = restored._persisted_fit_state_sidecar['predictive_datasets']['hrpt']
+    assert np.allclose(dataset['x'], np.asarray([1.0, 2.0]))
+    assert np.allclose(dataset['best_sample_prediction'], np.asarray([3.0, 4.0]))
+
+
+def test_read_analysis_results_sidecar_warns_when_expected_file_is_missing(tmp_path, monkeypatch):
+    from easydiffraction.io import results_sidecar as results_sidecar_mod
+
+    analysis = _analysis_with_predictive_sidecar()
+    warnings: list[str] = []
+    monkeypatch.setattr(results_sidecar_mod.log, 'warning', warnings.append)
+
+    results_sidecar_mod.read_analysis_results_sidecar(
+        analysis=analysis,
+        analysis_dir=Path(tmp_path),
+    )
+
+    assert analysis._persisted_fit_state_sidecar == {}
+    assert any('Expected Bayesian results sidecar is missing' in warning for warning in warnings)
+
+
+def test_sidecar_path_traversal_falls_back_to_local_results_file(tmp_path, monkeypatch):
+    from easydiffraction.io import results_sidecar as results_sidecar_mod
+
+    analysis_dir = Path(tmp_path) / 'analysis'
+    external_sidecar = Path(tmp_path) / 'outside.h5'
+    analysis = _analysis_with_predictive_sidecar()
+    analysis.bayesian_result._set_sidecar_file('../outside.h5')
+
+    warnings: list[str] = []
+    monkeypatch.setattr(results_sidecar_mod.log, 'warning', warnings.append)
+
+    results_sidecar_mod.write_analysis_results_sidecar(
+        analysis=analysis,
+        analysis_dir=analysis_dir,
+    )
+
+    assert (analysis_dir / 'results.h5').is_file()
+    assert not external_sidecar.exists()
+
+    restored = _analysis_with_predictive_sidecar()
+    restored.fit_results = None
+    restored.bayesian_result._set_sidecar_file('../outside.h5')
+    results_sidecar_mod.read_analysis_results_sidecar(
+        analysis=restored,
+        analysis_dir=analysis_dir,
+    )
+
+    assert 'predictive_datasets' in restored._persisted_fit_state_sidecar
+    assert any(
+        'Ignoring Bayesian sidecar file path outside the analysis directory' in warning
+        for warning in warnings
+    )
diff --git a/tests/unit/easydiffraction/project/categories/rendering/test_default.py b/tests/unit/easydiffraction/project/categories/rendering/test_default.py
index edb884272..07a168ae2 100644
--- a/tests/unit/easydiffraction/project/categories/rendering/test_default.py
+++ b/tests/unit/easydiffraction/project/categories/rendering/test_default.py
@@ -5,14 +5,18 @@
 
 
 def test_rendering_defaults():
+    from easydiffraction.display.plotting import PlotterEngineEnum
+    from easydiffraction.display.tables import TableEngineEnum
     from easydiffraction.project.categories.rendering.default import Rendering
 
     rendering = Rendering()
 
     assert rendering.type_info.tag == 'default'
     assert rendering._identity.category_code == 'rendering'
-    assert rendering.chart_engine.value == rendering.plotter.engine
-    assert rendering.table_engine.value == rendering.tabler.engine
+    assert rendering.chart_engine.value == 'auto'
+    assert rendering.table_engine.value == 'auto'
+    assert rendering.plotter.engine in [member.value for member in PlotterEngineEnum]
+    assert rendering.tabler.engine in [member.value for member in TableEngineEnum]
 
 
 def test_rendering_plotter_binds_parent():
@@ -28,6 +32,8 @@ def test_rendering_plotter_binds_parent():
 
 
 def test_rendering_setters_update_engines():
+    from easydiffraction.display.plotting import PlotterEngineEnum
+    from easydiffraction.display.tables import TableEngineEnum
     from easydiffraction.project.categories.rendering.default import Rendering
 
     rendering = Rendering()
@@ -40,6 +46,14 @@ def test_rendering_setters_update_engines():
     assert rendering.table_engine.value == 'rich'
     assert rendering.tabler.engine == 'rich'
 
+    rendering.chart_engine = 'auto'
+    rendering.table_engine = 'auto'
+
+    assert rendering.chart_engine.value == 'auto'
+    assert rendering.table_engine.value == 'auto'
+    assert rendering.plotter.engine == PlotterEngineEnum.default().value
+    assert rendering.tabler.engine == TableEngineEnum.default().value
+
 
 def test_rendering_from_cif_restores_types():
     from easydiffraction.project.categories.rendering.default import Rendering
diff --git a/tests/unit/easydiffraction/project/categories/verbosity/test_default.py b/tests/unit/easydiffraction/project/categories/verbosity/test_default.py
new file mode 100644
index 000000000..782bdbbf8
--- /dev/null
+++ b/tests/unit/easydiffraction/project/categories/verbosity/test_default.py
@@ -0,0 +1,44 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from __future__ import annotations
+
+import gemmi
+import pytest
+
+
+def test_verbosity_defaults_and_cif_output():
+    from easydiffraction.project.categories.verbosity.default import Verbosity
+
+    verbosity = Verbosity()
+
+    assert verbosity.type_info.tag == 'default'
+    assert verbosity._identity.category_code == 'verbosity'
+    assert verbosity.fit.value == 'full'
+    assert '_verbosity.fit full' in verbosity.as_cif
+
+
+def test_verbosity_setter_validates_enum_values():
+    from easydiffraction.project.categories.verbosity.default import Verbosity
+
+    verbosity = Verbosity()
+
+    verbosity.fit = 'short'
+    assert verbosity.fit.value == 'short'
+
+    verbosity.fit = 'silent'
+    assert verbosity.fit.value == 'silent'
+
+    with pytest.raises(ValueError, match="'verbose' is not a valid VerbosityEnum"):
+        verbosity.fit = 'verbose'
+
+
+def test_verbosity_from_cif_restores_fit_value():
+    from easydiffraction.project.categories.verbosity.default import Verbosity
+
+    verbosity = Verbosity()
+    block = gemmi.cif.read_string('data_test\n_verbosity.fit short\n').sole_block()
+
+    verbosity.from_cif(block)
+
+    assert verbosity.fit.value == 'short'
diff --git a/tests/unit/easydiffraction/project/categories/verbosity/test_factory.py b/tests/unit/easydiffraction/project/categories/verbosity/test_factory.py
new file mode 100644
index 000000000..3d3d6941b
--- /dev/null
+++ b/tests/unit/easydiffraction/project/categories/verbosity/test_factory.py
@@ -0,0 +1,25 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+
+from __future__ import annotations
+
+import pytest
+
+
+def test_verbosity_factory_default_and_create():
+    from easydiffraction.project.categories.verbosity.default import Verbosity
+    from easydiffraction.project.categories.verbosity.factory import VerbosityFactory
+
+    assert VerbosityFactory.default_tag() == 'default'
+    assert 'default' in VerbosityFactory.supported_tags()
+
+    verbosity = VerbosityFactory.create('default')
+
+    assert isinstance(verbosity, Verbosity)
+
+
+def test_verbosity_factory_rejects_unknown_tag():
+    from easydiffraction.project.categories.verbosity.factory import VerbosityFactory
+
+    with pytest.raises(ValueError, match=r"Unsupported type: 'missing'"):
+        VerbosityFactory.create('missing')
diff --git a/tests/unit/easydiffraction/project/test_display.py b/tests/unit/easydiffraction/project/test_display.py
index 1a892bca0..f534da493 100644
--- a/tests/unit/easydiffraction/project/test_display.py
+++ b/tests/unit/easydiffraction/project/test_display.py
@@ -47,12 +47,25 @@ def _recorder(*args, **kwargs):
         plot_calc=record('plot_calc'),
         plot_meas_vs_calc=record('plot_meas_vs_calc'),
         _plot_meas_vs_calc_request=record('_plot_meas_vs_calc_request'),
+        engine='plotly',
+        _resolve_x_axis=lambda expt_type, x: ('two_theta', 'two_theta', None, None, None),
     )
     project = SimpleNamespace(
-        analysis=SimpleNamespace(display=analysis_display),
+        analysis=SimpleNamespace(
+            display=analysis_display,
+            fit_results=SimpleNamespace(posterior_predictive={}),
+            bayesian_result=SimpleNamespace(
+                has_pair_cache=SimpleNamespace(value=False),
+                has_posterior_predictive=SimpleNamespace(value=False),
+            ),
+            bayesian_pair_caches=[],
+            bayesian_predictive_datasets=[],
+            _persisted_fit_state_sidecar={},
+        ),
         rendering=SimpleNamespace(plotter=plotter),
+        experiments={'hrpt': SimpleNamespace(type=SimpleNamespace())},
         free_parameters=[],
-        verbosity='full',
+        verbosity=SimpleNamespace(fit=SimpleNamespace(value='full')),
     )
     return project, calls
 
@@ -267,6 +280,67 @@ def fake_activity_indicator(label, *, verbosity):
     ]
 
 
+def test_posterior_predictive_skips_processing_indicator_for_restored_cache(monkeypatch):
+    import easydiffraction.project.display as display_mod
+
+    project, calls = _make_project_stub()
+    project.analysis = SimpleNamespace(
+        fit_results=object(),
+        bayesian_result=SimpleNamespace(has_posterior_predictive=SimpleNamespace(value=True)),
+        bayesian_predictive_datasets=[
+            SimpleNamespace(
+                experiment_name=SimpleNamespace(value='hrpt'),
+                x_axis_name=SimpleNamespace(value='two_theta'),
+                draws_path=SimpleNamespace(value=None),
+            )
+        ],
+        _persisted_fit_state_sidecar={
+            'predictive_datasets': {
+                'hrpt': {
+                    'x': [1.0, 2.0],
+                    'best_sample_prediction': [3.0, 4.0],
+                }
+            }
+        },
+    )
+    project.experiments = {'hrpt': SimpleNamespace(type=SimpleNamespace())}
+    project.rendering.plotter.engine = 'plotly'
+    project.rendering.plotter._resolve_x_axis = lambda expt_type, x: (
+        'two_theta',
+        'two_theta',
+        None,
+        None,
+        None,
+    )
+    display = ProjectDisplay(project)
+    indicator_calls: list[tuple[str, VerbosityEnum]] = []
+
+    @contextmanager
+    def fake_activity_indicator(label, *, verbosity):
+        indicator_calls.append((label, verbosity))
+        yield object()
+
+    monkeypatch.setattr(display_mod, 'activity_indicator', fake_activity_indicator)
+
+    display.posterior.predictive('hrpt')
+
+    assert calls == [
+        (
+            'plot_posterior_predictive',
+            (),
+            {
+                'expt_name': 'hrpt',
+                'style': 'band',
+                'x_min': None,
+                'x_max': None,
+                'show_residual': None,
+                'x': None,
+            },
+        )
+    ]
+    assert indicator_calls == []
+
+
 def test_posterior_distribution_without_param_plots_all_free_parameters():
     project, calls = _make_project_stub()
     project.free_parameters = ['a', 'b']
diff --git a/tests/unit/easydiffraction/project/test_project.py b/tests/unit/easydiffraction/project/test_project.py
index e3dcf7a19..3afbd363e 100644
--- a/tests/unit/easydiffraction/project/test_project.py
+++ b/tests/unit/easydiffraction/project/test_project.py
@@ -30,7 +30,7 @@ def test_project_verbosity_default():
     from easydiffraction.project.project import Project
 
     p = Project()
-    assert p.verbosity == 'full'
+    assert p.verbosity.fit.value == 'full'
 
 
 def test_project_verbosity_setter():
@@ -38,11 +38,11 @@ def test_project_verbosity_setter():
 
     p = Project()
     p.verbosity = 'short'
-    assert p.verbosity == 'short'
+    assert p.verbosity.fit.value == 'short'
     p.verbosity = 'silent'
-    assert p.verbosity == 'silent'
+    assert p.verbosity.fit.value == 'silent'
     p.verbosity = 'full'
-    assert p.verbosity == 'full'
+    assert p.verbosity.fit.value == 'full'
 
 
 def test_project_verbosity_invalid():
diff --git a/tests/unit/easydiffraction/project/test_project_config.py b/tests/unit/easydiffraction/project/test_project_config.py
index 3ba20eea1..be6576e94 100644
--- a/tests/unit/easydiffraction/project/test_project_config.py
+++ b/tests/unit/easydiffraction/project/test_project_config.py
@@ -23,8 +23,12 @@ def test_project_config_exposes_project_info_and_rendering_categories():
     assert config.info.path is None
     assert isinstance(config.info.created, datetime.datetime)
     assert isinstance(config.info.last_modified, datetime.datetime)
-    assert config.categories == [config.info, config.rendering]
-    assert config.parameters == config.info.parameters + config.rendering.parameters
+    assert config.verbosity._parent is config
+    assert config.verbosity.fit.value == 'full'
+    assert config.categories == [config.info, config.rendering, config.verbosity]
+    assert config.parameters == (
+        config.info.parameters + config.rendering.parameters + config.verbosity.parameters
+    )
 
 
 def test_project_config_as_cif_has_project_and_rendering_sections_without_data_header():
@@ -42,6 +46,29 @@ def test_project_config_as_cif_has_project_and_rendering_sections_without_data_h
     assert '_project.last_modified' in cif_text
     assert '_rendering.chart_engine' in cif_text
     assert '_rendering.table_engine' in cif_text
+    assert '_rendering.chart_engine auto' in cif_text
+    assert '_rendering.table_engine auto' in cif_text
+    assert '_verbosity.fit full' in cif_text
+
+
+def test_project_save_and_load_use_auto_rendering_defaults_when_unset(tmp_path):
+    from easydiffraction.project.project import Project
+
+    project = Project(name='beer', title='Beer title', description='Some description')
+    project.save_as(str(tmp_path / 'proj'))
+
+    project_cif = (tmp_path / 'proj' / 'project.cif').read_text()
+
+    assert not project_cif.startswith('data_')
+    assert '_rendering.chart_engine auto' in project_cif
+    assert '_rendering.table_engine auto' in project_cif
+    assert '_verbosity.fit full' in project_cif
+
+    loaded = Project.load(str(tmp_path / 'proj'))
+
+    assert loaded.rendering.chart_engine.value == 'auto'
+    assert loaded.rendering.table_engine.value == 'auto'
+    assert loaded.verbosity.fit.value == 'full'
 
 
 def test_project_save_and_load_keep_project_config_section_format(tmp_path):
@@ -57,6 +84,7 @@ def test_project_save_and_load_keep_project_config_section_format(tmp_path):
     assert '_project.id               beer' in project_cif
     assert '_rendering.chart_engine asciichartpy' in project_cif
     assert '_rendering.table_engine rich' in project_cif
+    assert '_verbosity.fit full' in project_cif
 
     loaded = Project.load(str(tmp_path / 'proj'))
     assert loaded.info.name == 'beer'
@@ -66,3 +94,36 @@ def test_project_save_and_load_keep_project_config_section_format(tmp_path):
     assert isinstance(loaded.info.last_modified, datetime.datetime)
     assert loaded.rendering.chart_engine.value == 'asciichartpy'
     assert loaded.rendering.table_engine.value == 'rich'
+    assert loaded.verbosity.fit.value == 'full'
+
+
+def test_project_save_wraps_long_description_as_cif_text_field(tmp_path):
+    from easydiffraction.project.project import Project
+
+    description = (
+        'This is the most minimal example of using EasyDiffraction. '
+        'It shows how to load a previously saved project from a directory '
+        'and run refinement in just a few lines of code.'
+    )
+    project = Project(name='beer', title='Beer title', description=description)
+    project.save_as(str(tmp_path / 'proj'))
+
+    project_cif = (tmp_path / 'proj' / 'project.cif').read_text()
+
+    assert '_project.description' in project_cif
+    description_tail = project_cif.split('_project.description', maxsplit=1)[1].lstrip(' ')
+    assert description_tail.startswith('\n;\n')
+    assert '\n;\n_project.created' in project_cif
+    description_block = description_tail.split('\n;\n', maxsplit=1)[1]
+    description_block = description_block.split('\n;\n_project.created', maxsplit=1)[0]
+    description_lines = description_block.splitlines()
+
+    assert len(description_lines) > 1
+    assert all(not line.startswith(';') for line in description_lines)
+    assert all(not line.endswith(';') for line in description_lines)
+    assert description_lines[0].startswith('This is the most minimal example')
+    assert description_lines[-1].endswith('lines of code.')
+
+    loaded = Project.load(str(tmp_path / 'proj'))
+
+    assert loaded.info.description == description
diff --git a/tests/unit/easydiffraction/project/test_project_load.py b/tests/unit/easydiffraction/project/test_project_load.py
index e44c12608..f19b4392d 100644
--- a/tests/unit/easydiffraction/project/test_project_load.py
+++ b/tests/unit/easydiffraction/project/test_project_load.py
@@ -125,6 +125,169 @@ def test_round_trips_constraints(self, tmp_path):
         assert loaded.analysis.constraints[0].expression.value == 'b_param = a_param'
         assert loaded.analysis.constraints.enabled is True
 
+    def test_round_trips_deterministic_fit_state_and_keeps_live_parameter_values(self, tmp_path):
+        original = Project(name='fit_state')
+        original.structures.create(name='lbco')
+        structure = original.structures['lbco']
+        structure.space_group.name_h_m = 'P m -3 m'
+        structure.cell.length_a = 3.88
+        parameter = structure.cell.length_a
+        parameter.free = True
+        parameter.uncertainty = 0.07
+        parameter.fit_min = 3.8
+        parameter.fit_max = 3.9
+        parameter._set_fit_bounds_uncertainty_multiplier(4.0)
+        parameter._fit_start_value = 3.87
+        parameter._fit_start_uncertainty = 0.02
+
+        original.analysis.fit_parameters.create(
+            param_unique_name=parameter.unique_name,
+            fit_min=parameter.fit_min,
+            fit_max=parameter.fit_max,
+            fit_bounds_uncertainty_multiplier=4.0,
+            start_value=3.87,
+            start_uncertainty=0.02,
+        )
+        original.analysis.fit_result._set_result_kind('deterministic')
+        original.analysis.fit_result._set_success(value=True)
+        original.analysis.fit_result._set_message('Fit converged')
+        original.analysis.fit_result._set_iterations(37)
+        original.analysis.fit_result._set_fitting_time(1.82)
+        original.analysis.fit_result._set_reduced_chi_square(1.031)
+        original.analysis.deterministic_result._set_optimizer_name('lmfit')
+        original.analysis.deterministic_result._set_method_name('leastsq')
+        original.analysis._set_has_persisted_fit_state(value=True)
+        original.save_as(str(tmp_path / 'proj'))
+
+        loaded = Project.load(str(tmp_path / 'proj'))
+        loaded_parameter = loaded.structures['lbco'].cell.length_a
+
+        assert loaded.analysis.fit_result.result_kind.value == 'deterministic'
+        assert loaded.analysis.fit_parameters[parameter.unique_name].fit_min.value == 3.8
+        assert loaded_parameter.value == 3.88
+        assert loaded_parameter.fit_min == 3.8
+        assert loaded_parameter.fit_max == 3.9
+        assert loaded_parameter.fit_bounds_uncertainty_multiplier == 4.0
+        assert loaded_parameter._fit_start_value == 3.87
+        assert loaded_parameter._fit_start_uncertainty == 0.02
+        assert loaded_parameter.uncertainty == 0.07
+
+    def test_round_trips_persisted_deterministic_correlation_summary_for_reloaded_display(
+        self,
+        tmp_path,
+    ):
+        from easydiffraction.display.plotting import Plotter
+
+        original = Project(name='fit_correlation_state')
+        original.structures.create(name='lbco')
+        structure = original.structures['lbco']
+        structure.space_group.name_h_m = 'P m -3 m'
+        structure.cell.length_a = 3.88
+        structure.cell.length_b = 3.89
+
+        parameter_a = structure.cell.length_a
+        parameter_b = structure.cell.length_b
+        for parameter, start_value in (
+            (parameter_a, 3.87),
+            (parameter_b, 3.88),
+        ):
+            parameter.free = True
+            parameter.uncertainty = 0.05
+            parameter.fit_min = 3.8
+            parameter.fit_max = 3.9
+            parameter._set_fit_bounds_uncertainty_multiplier(4.0)
+            parameter._fit_start_value = start_value
+            parameter._fit_start_uncertainty = 0.02
+            original.analysis.fit_parameters.create(
+                param_unique_name=parameter.unique_name,
+                fit_min=parameter.fit_min,
+                fit_max=parameter.fit_max,
+                fit_bounds_uncertainty_multiplier=4.0,
+                start_value=start_value,
+                start_uncertainty=0.02,
+            )
+
+        original.analysis.fit_result._set_result_kind('deterministic')
+        original.analysis.fit_result._set_success(value=True)
+        original.analysis.fit_result._set_message('Fit converged')
+        original.analysis.fit_result._set_iterations(21)
+        original.analysis.fit_result._set_fitting_time(0.74)
+        original.analysis.fit_result._set_reduced_chi_square(1.031)
+        original.analysis.deterministic_result._set_optimizer_name('lmfit')
+        original.analysis.deterministic_result._set_method_name('leastsq')
+        original.analysis.deterministic_result._set_objective_name('chi-square')
+        original.analysis.deterministic_result._set_objective_value(1.031)
+        original.analysis.deterministic_result._set_n_data_points(120)
+        original.analysis.deterministic_result._set_n_parameters(2)
+        original.analysis.deterministic_result._set_n_free_parameters(2)
+        original.analysis.deterministic_result._set_degrees_of_freedom(118)
+        original.analysis.deterministic_result._set_covariance_available(value=False)
+        original.analysis.deterministic_result._set_correlation_available(value=True)
+        original.analysis.fit_parameter_correlations.create(
+            source_kind='deterministic',
+            param_unique_name_i=parameter_b.unique_name,
+            param_unique_name_j=parameter_a.unique_name,
+            correlation=0.42,
+        )
+        original.analysis._set_has_persisted_fit_state(value=True)
+        original.save_as(str(tmp_path / 'proj'))
+
+        loaded = Project.load(str(tmp_path / 'proj'))
+        plotter = Plotter()
+        plotter._set_project(loaded)
+
+        corr_df = plotter._get_param_correlation_dataframe()
+
+        assert corr_df is not None
+        assert list(corr_df.index) == [parameter_a.unique_name, parameter_b.unique_name]
+        assert list(corr_df.columns) == [parameter_a.unique_name, parameter_b.unique_name]
+        assert corr_df.loc[parameter_a.unique_name, parameter_a.unique_name] == pytest.approx(1.0)
+        assert corr_df.loc[parameter_b.unique_name, parameter_b.unique_name] == pytest.approx(1.0)
+        assert corr_df.loc[parameter_a.unique_name, parameter_b.unique_name] == pytest.approx(0.42)
+        assert corr_df.loc[parameter_b.unique_name, parameter_a.unique_name] == pytest.approx(0.42)
+
+    def test_round_trips_bayesian_sampler_settings_to_live_dream_minimizer(self, tmp_path):
+        original = Project(name='bayes_state')
+        original.analysis.fitting.minimizer_type = 'bumps (dream)'
+        original.analysis.fit_result._set_result_kind('bayesian')
+        original.analysis.bayesian_sampler._set_steps(300)
+        original.analysis.bayesian_sampler._set_burn(60)
+        original.analysis.bayesian_sampler._set_thin(2)
+        original.analysis.bayesian_sampler._set_pop(8)
+        original.analysis.bayesian_sampler._set_parallel(0)
+        original.analysis.bayesian_sampler._set_init('lhs')
+        original.analysis._set_has_persisted_fit_state(value=True)
+        original.save_as(str(tmp_path / 'proj'))
+
+        loaded = Project.load(str(tmp_path / 'proj'))
+        minimizer = loaded.analysis.fitting.minimizer
+
+        assert minimizer is not None
+        assert minimizer.steps == 300
+        assert minimizer.burn == 60
+        assert minimizer.thin == 2
+        assert minimizer.pop == 8
+        assert minimizer.parallel == 0
+        assert minimizer.init.value == 'lhs'
+
+    def test_round_trips_legacy_bayesian_steps_and_burn_to_live_dream_minimizer(self, tmp_path):
+        original = Project(name='legacy_bayes_state')
+        original.analysis.fitting.minimizer_type = 'bumps (dream)'
+        original.analysis.fit_result._set_result_kind('bayesian')
+        original.analysis.bayesian_sampler._set_steps(300)
+        original.analysis.bayesian_sampler._set_burn(60)
+        original.analysis._set_has_persisted_fit_state(value=True)
+        original.save_as(str(tmp_path / 'proj'))
+
+        loaded = Project.load(str(tmp_path / 'proj'))
+        minimizer = loaded.analysis.fitting.minimizer
+
+        assert minimizer is not None
+        assert minimizer.steps == 300
+        assert minimizer.burn == 60
+        assert minimizer.thin == 1
+        assert minimizer.pop == 4
+
 
 class TestLoadAnalysisCifFallback:
     """Load falls back from analysis/analysis.cif to analysis.cif at root."""
diff --git a/tests/unit/easydiffraction/project/test_project_save.py b/tests/unit/easydiffraction/project/test_project_save.py
index 27212251b..ff8a9b9be 100644
--- a/tests/unit/easydiffraction/project/test_project_save.py
+++ b/tests/unit/easydiffraction/project/test_project_save.py
@@ -62,3 +62,62 @@ def test_project_save_lists_existing_analysis_results_csv(tmp_path, monkeypatch,
     out = capsys.readouterr().out
     assert 'analysis.cif' in out
     assert 'results.csv' in out
+
+
+def test_project_save_as_overwrites_existing_directory_by_default(tmp_path, monkeypatch):
+    from easydiffraction.analysis.analysis import Analysis
+    from easydiffraction.project.project import Project
+    from easydiffraction.project.project_info import ProjectInfo
+    from easydiffraction.summary.summary import Summary
+
+    monkeypatch.setattr(ProjectInfo, 'as_cif', property(lambda self: 'info'))
+    monkeypatch.setattr(Analysis, 'as_cif', property(lambda self: 'analysis'))
+    monkeypatch.setattr(Summary, 'as_cif', lambda self: 'summary')
+
+    target = tmp_path / 'proj_dir'
+    stale_file = target / 'stale.txt'
+    target.mkdir()
+    stale_file.write_text('stale')
+
+    project = Project(name='p1')
+    project.save_as(str(target))
+
+    assert not stale_file.exists()
+    assert (target / 'project.cif').is_file()
+
+
+def test_project_save_as_preserves_existing_directory_when_disabled(tmp_path, monkeypatch):
+    from easydiffraction.analysis.analysis import Analysis
+    from easydiffraction.project.project import Project
+    from easydiffraction.project.project_info import ProjectInfo
+    from easydiffraction.summary.summary import Summary
+
+    monkeypatch.setattr(ProjectInfo, 'as_cif', property(lambda self: 'info'))
+    monkeypatch.setattr(Analysis, 'as_cif', property(lambda self: 'analysis'))
+    monkeypatch.setattr(Summary, 'as_cif', lambda self: 'summary')
+
+    target = tmp_path / 'proj_dir'
+    stale_file = target / 'stale.txt'
+    target.mkdir()
+    stale_file.write_text('stale')
+
+    project = Project(name='p1')
+    project.save_as(
+        str(target),
+        overwrite=False,
+    )
+
+    assert stale_file.exists()
+    assert (target / 'project.cif').is_file()
+
+
+def test_project_save_omits_empty_fit_state_sections(tmp_path):
+    from easydiffraction.project.project import Project
+
+    project = Project(name='no_fit_state')
+    project.save_as(str(tmp_path / 'proj'))
+
+    analysis_cif = (tmp_path / 'proj' / 'analysis' / 'analysis.cif').read_text()
+
+    assert '_fit_parameter.param_unique_name' not in analysis_cif
+    assert '_fit_result.result_kind' not in analysis_cif
diff --git a/tests/unit/easydiffraction/test___main__.py b/tests/unit/easydiffraction/test___main__.py
index 62ad8c68c..a78710b35 100644
--- a/tests/unit/easydiffraction/test___main__.py
+++ b/tests/unit/easydiffraction/test___main__.py
@@ -44,6 +44,14 @@ def test_cli_subcommands_call_utils(monkeypatch):
     import easydiffraction.__main__ as main_mod
 
     logs = []
+    monkeypatch.setattr(ed, 'list_data', lambda: logs.append('LIST_DATA'))
+    monkeypatch.setattr(
+        ed,
+        'download_data',
+        lambda id, destination='data', overwrite=False: logs.append(
+            f'DATA_{id}_{destination}_{overwrite}'
+        ),
+    )
     monkeypatch.setattr(ed, 'list_tutorials', lambda: logs.append('LIST'))
     monkeypatch.setattr(
         ed,
@@ -56,14 +64,25 @@ def test_cli_subcommands_call_utils(monkeypatch):
         lambda id, destination='tutorials', overwrite=False: logs.append(f'DOWNLOAD_{id}'),
     )
 
-    res1 = runner.invoke(main_mod.app, ['list-tutorials'])
-    res2 = runner.invoke(main_mod.app, ['download-all-tutorials'])
-    res3 = runner.invoke(main_mod.app, ['download-tutorial', '1'])
+    res0 = runner.invoke(main_mod.app, ['list-data'])
+    res1 = runner.invoke(main_mod.app, ['download-data', '30', '--destination', 'projects'])
+    res2 = runner.invoke(main_mod.app, ['list-tutorials'])
+    res3 = runner.invoke(main_mod.app, ['download-all-tutorials'])
+    res4 = runner.invoke(main_mod.app, ['download-tutorial', '1'])
 
+    assert res0.exit_code == 0
     assert res1.exit_code == 0
     assert res2.exit_code == 0
     assert res3.exit_code == 0
-    assert logs == ['LIST', 'DOWNLOAD_ALL', 'DOWNLOAD_1']
+    assert res4.exit_code == 0
+    assert logs == ['LIST_DATA', 'DATA_30_projects_False', 'LIST', 'DOWNLOAD_ALL', 'DOWNLOAD_1']
+
+
+def test_cli_project_first_argument_normalization_supports_global_data_commands():
+    import easydiffraction.__main__ as main_mod
+
+    assert main_mod._normalized_cli_args(['list-data']) == ['list-data']
+    assert main_mod._normalized_cli_args(['download-data', '30']) == ['download-data', '30']
 
 
 def test_cli_fit_loads_and_fits(monkeypatch, tmp_path):
diff --git a/tests/unit/easydiffraction/utils/test_environment.py b/tests/unit/easydiffraction/utils/test_environment.py
index 691b0a9c5..fab45e723 100644
--- a/tests/unit/easydiffraction/utils/test_environment.py
+++ b/tests/unit/easydiffraction/utils/test_environment.py
@@ -84,3 +84,44 @@ def test_can_use_ipython_display_with_none(self):
         from easydiffraction.utils.environment import can_use_ipython_display
 
         assert can_use_ipython_display(None) is False
+
+
+class TestArtifactPaths:
+    def test_resolve_artifact_path_uses_env_root(self, monkeypatch, tmp_path):
+        from easydiffraction.utils.environment import resolve_artifact_path
+
+        monkeypatch.setenv('EASYDIFFRACTION_ARTIFACT_ROOT', 'tmp/tutorials')
+        monkeypatch.setenv('PIXI_PROJECT_ROOT', str(tmp_path))
+
+        assert resolve_artifact_path('data') == tmp_path / 'tmp' / 'tutorials' / 'data'
+
+    def test_resolve_artifact_path_uses_tutorial_fallback(self, monkeypatch, tmp_path):
+        import easydiffraction.utils.environment as env
+
+        repo_root = tmp_path / 'repo'
+        tutorials_dir = repo_root / 'docs' / 'docs' / 'tutorials'
+        tutorials_dir.mkdir(parents=True)
+
+        monkeypatch.delenv('EASYDIFFRACTION_ARTIFACT_ROOT', raising=False)
+        monkeypatch.delenv('PIXI_PROJECT_ROOT', raising=False)
+        monkeypatch.chdir(tutorials_dir)
+        monkeypatch.setattr(env, '_repo_root', lambda: repo_root)
+
+        assert env.resolve_artifact_path('data') == repo_root / 'tmp' / 'tutorials' / 'data'
+
+    def test_create_artifact_temp_dir_uses_tutorial_fallback(self, monkeypatch, tmp_path):
+        import easydiffraction.utils.environment as env
+
+        repo_root = tmp_path / 'repo'
+        tutorials_dir = repo_root / 'docs' / 'docs' / 'tutorials'
+        tutorials_dir.mkdir(parents=True)
+
+        monkeypatch.delenv('EASYDIFFRACTION_ARTIFACT_ROOT', raising=False)
+        monkeypatch.delenv('PIXI_PROJECT_ROOT', raising=False)
+        monkeypatch.chdir(tutorials_dir)
+        monkeypatch.setattr(env, '_repo_root', lambda: repo_root)
+
+        created_dir = env.create_artifact_temp_dir('ed_zip_')
+
+        assert created_dir.is_dir()
+        assert created_dir.parent == repo_root / 'tmp' / 'tutorials'
diff --git a/tests/unit/easydiffraction/utils/test_utils_coverage.py b/tests/unit/easydiffraction/utils/test_utils_coverage.py
index 9357e2725..d2d07159e 100644
--- a/tests/unit/easydiffraction/utils/test_utils_coverage.py
+++ b/tests/unit/easydiffraction/utils/test_utils_coverage.py
@@ -433,6 +433,42 @@ def test_download_data_no_description(monkeypatch, tmp_path, capsys):
     assert 'Data #1' in out
 
 
+def test_download_data_uses_tutorial_artifact_root_fallback(monkeypatch, tmp_path):
+    import easydiffraction.utils.environment as env
+    import easydiffraction.utils.utils as MUT
+
+    repo_root = tmp_path / 'repo'
+    tutorials_dir = repo_root / 'docs' / 'docs' / 'tutorials'
+    tutorials_dir.mkdir(parents=True)
+
+    fake_index = {
+        '1': {
+            'path': 'data.xye',
+            'hash': None,
+            'description': 'Test data',
+        }
+    }
+    monkeypatch.setattr(MUT, '_fetch_data_index', lambda: fake_index)
+    monkeypatch.setattr(env, '_repo_root', lambda: repo_root)
+    monkeypatch.delenv('EASYDIFFRACTION_ARTIFACT_ROOT', raising=False)
+    monkeypatch.delenv('PIXI_PROJECT_ROOT', raising=False)
+    monkeypatch.chdir(tutorials_dir)
+
+    def fake_retrieve(url, known_hash, fname, path):
+        import pathlib
+
+        pathlib.Path(path, fname).write_text('x y e')
+        return str(pathlib.Path(path, fname))
+
+    monkeypatch.setattr(MUT.pooch, 'retrieve', fake_retrieve)
+
+    result = MUT.download_data(id=1, destination='data')
+
+    expected_path = repo_root / 'tmp' / 'tutorials' / 'data' / 'ed-1.xye'
+    assert result == str(expected_path)
+    assert expected_path.exists()
+
+
 # --- download_tutorial with overwrite=True ------------------------------------
 
 
diff --git a/tests/unit/test_benchmark_tutorials.py b/tests/unit/test_benchmark_tutorials.py
new file mode 100644
index 000000000..c0b390ef7
--- /dev/null
+++ b/tests/unit/test_benchmark_tutorials.py
@@ -0,0 +1,133 @@
+# SPDX-FileCopyrightText: 2026 EasyScience contributors <https://github.com/easyscience>
+# SPDX-License-Identifier: BSD-3-Clause
+"""Unit tests for tutorial benchmark CSV persistence."""
+
+from __future__ import annotations
+
+import csv
+import importlib.util
+import sys
+from argparse import Namespace
+from pathlib import Path
+from types import SimpleNamespace
+
+
+def _load_module():
+    module_path = Path(__file__).resolve().parents[2] / 'tools' / 'benchmark_tutorials.py'
+    spec = importlib.util.spec_from_file_location('benchmark_tutorials', module_path)
+    assert spec is not None
+    assert spec.loader is not None
+    module = importlib.util.module_from_spec(spec)
+    sys.modules[spec.name] = module
+    spec.loader.exec_module(module)
+    return module
+
+
+MUT = _load_module()
+
+
+def test_append_result_writes_one_row(tmp_path):
+    output_path = tmp_path / 'benchmark.csv'
+
+    MUT._write_csv_header(output_path)
+    MUT._append_result(
+        output_path,
+        MUT.TutorialBenchmarkResult(
+            tutorial_name='ed-21.py',
+            elapsed_seconds=12.3456,
+            status='ok',
+            return_code=0,
+        ),
+    )
+
+    with output_path.open(encoding='utf-8', newline='') as handle:
+        rows = list(csv.DictReader(handle))
+
+    assert rows == [
+        {
+            'tutorial_name': 'ed-21.py',
+            'elapsed_seconds': '12.346',
+            'status': 'ok',
+            'return_code': '0',
+        }
+    ]
+
+
+def test_main_appends_first_result_before_second_tutorial_starts(monkeypatch, tmp_path):
+    tutorial_dir = tmp_path / 'tutorials'
+    tutorial_dir.mkdir()
+    first_tutorial = tutorial_dir / 'ed-01.py'
+    second_tutorial = tutorial_dir / 'ed-02.py'
+    first_tutorial.write_text('print("first")\n', encoding='utf-8')
+    second_tutorial.write_text('print("second")\n', encoding='utf-8')
+
+    output_path = tmp_path / 'benchmarking' / 'results.csv'
+    args = Namespace(
+        tutorial_dir=tutorial_dir,
+        output_dir=tmp_path / 'benchmarking',
+        pattern=[],
+    )
+
+    monkeypatch.setattr(
+        MUT,
+        'build_parser',
+        lambda: SimpleNamespace(parse_args=lambda: args),
+    )
+    monkeypatch.setattr(MUT, '_build_output_path', lambda output_dir: output_path)
+    monkeypatch.setattr(MUT, '_build_env', dict)
+
+    def fake_run_tutorial(
+        script_path: Path,
+        tutorial_dir_path: Path,
+        env: dict[str, str],
+    ) -> MUT.TutorialBenchmarkResult:
+        del tutorial_dir_path, env
+        if script_path == first_tutorial:
+            with output_path.open(encoding='utf-8', newline='') as handle:
+                rows = list(csv.reader(handle))
+            assert rows == [MUT.CSV_HEADER]
+            return MUT.TutorialBenchmarkResult(
+                tutorial_name='ed-01.py',
+                elapsed_seconds=1.0,
+                status='ok',
+                return_code=0,
+            )
+
+        with output_path.open(encoding='utf-8', newline='') as handle:
+            rows = list(csv.DictReader(handle))
+        assert rows == [
+            {
+                'tutorial_name': 'ed-01.py',
+                'elapsed_seconds': '1.000',
+                'status': 'ok',
+                'return_code': '0',
+            }
+        ]
+        return MUT.TutorialBenchmarkResult(
+            tutorial_name='ed-02.py',
+            elapsed_seconds=2.0,
+            status='ok',
+            return_code=0,
+        )
+
+    monkeypatch.setattr(MUT, '_run_tutorial', fake_run_tutorial)
+
+    exit_code = MUT.main()
+
+    assert exit_code == 0
+    with output_path.open(encoding='utf-8', newline='') as handle:
+        rows = list(csv.DictReader(handle))
+    assert rows == [
+        {
+            'tutorial_name': 'ed-01.py',
+            'elapsed_seconds': '1.000',
+            'status': 'ok',
+            'return_code': '0',
+        },
+        {
+            'tutorial_name': 'ed-02.py',
+            'elapsed_seconds': '2.000',
+            'status': 'ok',
+            'return_code': '0',
+        },
+    ]
diff --git a/tools/benchmark_tutorials.py b/tools/benchmark_tutorials.py
new file mode 100644
index 000000000..e28d77e98
--- /dev/null
+++ b/tools/benchmark_tutorials.py
@@ -0,0 +1,216 @@
+"""Benchmark tutorial scripts and save timing results."""
+
+from __future__ import annotations
+
+import argparse
+import csv
+import os
+import platform
+import subprocess  # noqa: S404
+import sys
+import time
+from dataclasses import dataclass
+from datetime import datetime
+from pathlib import Path
+from pathlib import PurePosixPath
+
+ROOT = Path(__file__).resolve().parents[1]
+SRC_ROOT = ROOT / 'src'
+DEFAULT_TUTORIAL_DIR = ROOT / 'docs' / 'docs' / 'tutorials'
+DEFAULT_OUTPUT_DIR = ROOT / 'docs' / 'dev' / 'benchmarking'
+CHECKPOINT_DIR_NAME = '.ipynb_checkpoints'
+CSV_HEADER = ['tutorial_name', 'elapsed_seconds', 'status', 'return_code']
+
+
+@dataclass(frozen=True)
+class TutorialBenchmarkResult:
+    """Store timing data for one tutorial run."""
+
+    tutorial_name: str
+    elapsed_seconds: float
+    status: str
+    return_code: int
+
+
+def _relative_display_path(path: Path, start_path: Path) -> str:
+    try:
+        return path.relative_to(start_path).as_posix()
+    except ValueError:
+        return path.as_posix()
+
+
+def _slugify(value: str) -> str:
+    return value.lower().replace(' ', '-').replace('/', '-')
+
+
+def _build_env() -> dict[str, str]:
+    env = os.environ.copy()
+    if SRC_ROOT.exists():
+        existing_pythonpath = env.get('PYTHONPATH', '')
+        env['PYTHONPATH'] = (
+            str(SRC_ROOT)
+            if not existing_pythonpath
+            else str(SRC_ROOT) + os.pathsep + existing_pythonpath
+        )
+    return env
+
+
+def _discover_tutorials(tutorial_dir: Path) -> list[Path]:
+    return [
+        path
+        for path in sorted(tutorial_dir.rglob('*.py'))
+        if CHECKPOINT_DIR_NAME not in path.parts
+    ]
+
+
+def _matches_requested_patterns(
+    script_path: Path,
+    tutorial_dir: Path,
+    patterns: list[str],
+) -> bool:
+    if not patterns:
+        return True
+
+    rel_path = PurePosixPath(_relative_display_path(script_path, tutorial_dir))
+    return any(rel_path.match(pattern) or script_path.name == pattern for pattern in patterns)
+
+
+def _run_tutorial(
+    script_path: Path,
+    tutorial_dir: Path,
+    env: dict[str, str],
+) -> TutorialBenchmarkResult:
+    tutorial_name = _relative_display_path(script_path, tutorial_dir)
+    start_time = time.perf_counter()
+    result = subprocess.run(  # noqa: S603
+        [sys.executable, str(script_path)],
+        cwd=str(ROOT),
+        env=env,
+        capture_output=True,
+        text=True,
+        encoding='utf-8',
+    )
+    elapsed_seconds = time.perf_counter() - start_time
+    status = 'ok' if result.returncode == 0 else 'failed'
+
+    if result.returncode == 0:
+        print(f'        OK      {elapsed_seconds:.1f}s')
+    else:
+        print(f'        FAILED  {elapsed_seconds:.1f}s', file=sys.stderr)
+        details = ((result.stdout or '') + (result.stderr or '')).strip()
+        if details:
+            print(details, file=sys.stderr)
+
+    return TutorialBenchmarkResult(
+        tutorial_name=tutorial_name,
+        elapsed_seconds=elapsed_seconds,
+        status=status,
+        return_code=result.returncode,
+    )
+
+
+def _build_output_path(output_dir: Path) -> Path:
+    timestamp = datetime.now().astimezone().strftime('%Y%m%d-%H%M%S')
+    system_name = _slugify(platform.system())
+    machine_name = _slugify(platform.machine())
+    python_name = f'py{sys.version_info.major}{sys.version_info.minor}'
+    file_name = (
+        f'{timestamp}_{system_name}-{machine_name}_'
+        f'{python_name}_tutorial-benchmarks.csv'
+    )
+    return output_dir / file_name
+
+
+def _write_csv_header(output_path: Path) -> None:
+    output_path.parent.mkdir(parents=True, exist_ok=True)
+    with output_path.open('w', encoding='utf-8', newline='') as handle:
+        writer = csv.writer(handle)
+        writer.writerow(CSV_HEADER)
+
+
+def _append_result(output_path: Path, result: TutorialBenchmarkResult) -> None:
+    with output_path.open('a', encoding='utf-8', newline='') as handle:
+        writer = csv.writer(handle)
+        writer.writerow(
+            [
+                result.tutorial_name,
+                f'{result.elapsed_seconds:.3f}',
+                result.status,
+                result.return_code,
+            ]
+        )
+
+
+def build_parser() -> argparse.ArgumentParser:
+    parser = argparse.ArgumentParser(
+        description='Run tutorial scripts sequentially and record timings.',
+    )
+    parser.add_argument(
+        '--tutorial-dir',
+        type=Path,
+        default=DEFAULT_TUTORIAL_DIR,
+        help='Directory containing tutorial scripts to benchmark.',
+    )
+    parser.add_argument(
+        '--output-dir',
+        type=Path,
+        default=DEFAULT_OUTPUT_DIR,
+        help='Directory where the benchmark CSV should be written.',
+    )
+    parser.add_argument(
+        '--pattern',
+        action='append',
+        default=[],
+        help=(
+            'Glob for tutorial paths relative to the tutorial directory. '
+            'Pass multiple times to benchmark a subset.'
+        ),
+    )
+    return parser
+
+
+def main() -> int:
+    args = build_parser().parse_args()
+    tutorial_dir = args.tutorial_dir.resolve()
+    output_dir = args.output_dir.resolve()
+
+    if not tutorial_dir.is_dir():
+        print(f'Tutorial directory not found: {tutorial_dir}', file=sys.stderr)
+        return 1
+
+    tutorials = [
+        path
+        for path in _discover_tutorials(tutorial_dir)
+        if _matches_requested_patterns(path, tutorial_dir, args.pattern)
+    ]
+    if not tutorials:
+        print('No tutorial scripts matched the requested pattern(s).', file=sys.stderr)
+        return 1
+
+    output_path = _build_output_path(output_dir)
+    _write_csv_header(output_path)
+
+    env = _build_env()
+    results: list[TutorialBenchmarkResult] = []
+    for index, tutorial_path in enumerate(tutorials, start=1):
+        tutorial_name = _relative_display_path(tutorial_path, tutorial_dir)
+        print(f'[{index:2}/{len(tutorials)}] Running {tutorial_name}')
+        result = _run_tutorial(tutorial_path, tutorial_dir, env)
+        results.append(result)
+        _append_result(output_path, result)
+
+    total_elapsed = sum(result.elapsed_seconds for result in results)
+    failure_count = sum(result.status == 'failed' for result in results)
+
+    print(f'Wrote benchmark results to {_relative_display_path(output_path, ROOT)}')
+    print(f'Total elapsed time: {total_elapsed:.3f}s')
+
+    if failure_count:
+        print(f'Failed tutorials: {failure_count}', file=sys.stderr)
+        return 1
+
+    return 0
+
+
+if __name__ == '__main__':
+    raise SystemExit(main())
\ No newline at end of file