AdaWorldAPI · AdaWorldAPI · Jul 18, 2026 · Jul 18, 2026 · Jul 18, 2026
diff --git a/.claude/board/AGENT_LOG.md b/.claude/board/AGENT_LOG.md
@@ -1,3 +1,12 @@
+## 2026-07-18 — GraphRAG representations inventory — 7 papers × V3 substrate matrix (6 Opus paper-readers + 1 Opus v3-harvest; main-thread synthesis)
+
+- **Task:** operator asked for an inventory of 7 papers (6 arXiv + MDPI), formulate 8 representations plus the v3 "should-have-built" set, and answer a matrix per representation (format / witness-ref / witness / context / basins / vertical-horizontal-vs-edges / time / NARS / causality-trajectory-candidate / wire) + "which probe considered all tenants in every SoA."
+- **Fleet:** 6 background Opus readers (SAP-Practical, AriGraph fidelity-audit, PersonalAI, GraphRAG-under-Fire, GraphRAG-FI, StepChain) + MDPI WebFetch + 1 Opus v3-harvest (facet ladder L1-L8, 13-tenant catalogue, ValueSchema, should-have-built backlog). Main thread synthesized.
+- **Deliverable:** `.claude/knowledge/graphrag-representations-inventory.md` — the 7-paper inventory, the facet ladder, the 13(+1) tenants, THE MATRIX (16 representations), the causality-trajectory subset, the all-tenants-in-one-SoA verdict, priority wire order.
+- **Probe answer:** "all tenants in every SoA" = `PROBE preset-vs-dispatch` (EPIPHANIES:1184) → `ValueSchema::Full` ("every tenant", compile-proven `canonical_node.rs:1127`) + the `ocr.rs:104-115` write-path-as-function-of-live-tenants dispatch.
+- **Headline gaps (paper-surfaced, unbuilt):** RRF fusion (keystone), provenance→trust wire, chained episodic_search (Eq.1 — the crate is "AriGraph in name, RAG in retrieval"), evidence-chain `Πsᵤ`, thesis partition, sub-question decomposition, BeamSearch. GraphRAG-FI + PersonalAI independently re-derive the MUL/free-energy + preserve-contradictions stances.
+- **Commit / Tests / Outcome:** commit `36327b0`; docs-only (no code / build surface → no tests run); CodeRabbit 5/5 pre-merge checks green + 4 minor doc-consistency comments addressed in a follow-up commit. Outcome: inventory + 16-representation matrix + probe answer shipped; RRF / provenance→trust / episodic_search (Eq.1) named as follow-up wires. Branch `claude/happy-hamilton-0azlw4` (restarted from clean post-#722 main); PR #723.
+
 ## 2026-07-18 — EpisodicMemory::basins() — usable episodic-witness basin partition (operator directive: "usable code, not paper churn") — main thread
 
 - **Task:** the operator asked for the AriGraph episodic-witness basins as USABLE CODE (not a probe, not jc-coupled). AriGraph had `communities()` (structural, #714) but no experiential partition (`grep basin` → nothing). Built the complement.

diff --git a/.claude/board/EPIPHANIES.md b/.claude/board/EPIPHANIES.md
@@ -1,3 +1,17 @@
+## 2026-07-18 — E-GRAPHRAG-REPRESENTATIONS-INVENTORY-1 — 7 GraphRAG papers map cleanly onto the V3 substrate; the crate is "AriGraph in name, RAG in retrieval"; the retrieval keystone (RRF) and the causality-trajectory wires are one-function-short gaps
+
+**Status:** SYNTHESIS (2026-07-18, `.claude/knowledge/graphrag-representations-inventory.md`). Grounded — paper briefs from 6 Opus readers + MDPI; substrate rows source-verified against `canonical_node.rs`/`class_view.rs`/`facet.rs`/`ocr.rs`. Honesty flags [G]/[H]/proposed carried through.
+
+**The finding:** all 7 papers (AriGraph, PersonalAI, Document-GraphRAG, SAP-Practical, StepChain, GraphRAG-FI, GraphRAG-under-Fire) cluster onto existing V3 carriers — none needs a new layer. (1) **AriGraph fidelity gap:** the `arigraph/` crate has the update-rule + agnostic-graph but runs a *parallel Hamming top-k* for episodic recall = the RAG baseline AriGraph beats; the first-class episodic edge (`Ee` / the unwired `EpisodicEdges64`) + the chained semantic→episodic search (their Eq.1) are MISSING. (2) **The retrieval keystone (RRF fusion) is one free fn short** — every ranked leg (BM25/PPR/CAM-PQ) exists, no combinator; the G0 harness compares by rank position instead of fusing. (3) **Provenance→trust is unwired** — `WitnessCorpus.source_url` exists but never reaches `TruthValue.confidence` (the only defense that works vs GRAGPOISON, and load-bearing for OCR mixed-trust docs regardless of adversary). (4) **GraphRAG-FI + PersonalAI independently re-derive** the workspace's MUL/free-energy meta-uncertainty and preserve-contradictions stances — validation, not novelty. (5) **PersonalAI's `thesis` vertex** (per-proposition grouping, their most load-bearing memory type) is a third partition the workspace lacks — a community-within-episode, parallel to `EpisodicBasins`.
+
+**Causality-trajectory subset** (owned + witnessed + NARS-revised via context edges): CausalEdge64/MaterializedEdges (the particle), the temporal.rs stream / Lance-versions (the carrier), SPO-with-truth + episodic-witness edge + evidence-chain `Πsᵤ` (candidates); basins GROUND, palette/COCA/fingerprint are not trajectories.
+
+**All-tenants-in-one-SoA:** `PROBE preset-vs-dispatch` (E-1184) → `ValueSchema::Full` = every tenant (compile-proven `canonical_node.rs:1127`); the `ocr.rs:104-115` `schema.has(tenant){write lane}` dispatch is the mechanism — representations are tenant lanes / ClassView readings of ONE SoA, re-resolved from the classid, no boundary crossing.
+
+**Reading A (operator, this session) — PROPOSED, not sanctioned:** `2×(basin:identity SPO)` grouping of the L1–L4 6×(8:8) plane (`[semantic SPO | witness SPO]`) closes the episodic-edge gap in-register; falsifier = per-typed-plane basin-byte agreement with `communities()`/`basins()` (never pooled — the corrected S1). [probe-gated]
+
+**Cross-ref:** `graphrag-doc-retrieval-soa-integration-v1.md` (the D-GR-* plan these wire into), `context-role-traversal-tissue.md` (the register polymorphism), `E-MARKOV-TEMPORAL-STREAM-1` (temporal wave), `E-THINKING-SPINE-CHESS-EVIDENCE-1` (basins GROUND / SPO ANCHOR / WAVE IDENTIFIES), `E-EMPIRICAL-VS-SCIENTIFIC-JC-1` (probes consume jc, never couple).
+
 ## 2026-07-18 — E-S1-PROBE-REMOVED-1 — the S1 community-basin probe and the lance-graph `jc` dev-coupling are REMOVED. The probe pooled ~1450 typed relation planes into one undirected similarity graph and read a scalar φ as a "community-id earns the mint" verdict — a flattening artifact. jc is the scientific crate; this empirical probe does not belong coupled to it.
 
 **Status:** SHIPPED (deletes `crates/lance-graph/examples/p_community_basin_agree.rs`; removes the `jc` dev-dependency from lance-graph `Cargo.toml` + `Cargo.lock`. The #722-only real-corpus variant + its schema.org data blob + python extractor never merged and are dropped by resetting the branch to clean main.)

diff --git a/.claude/knowledge/graphrag-representations-inventory.md b/.claude/knowledge/graphrag-representations-inventory.md
@@ -0,0 +1,156 @@
+# GraphRAG Representations Inventory — 7 papers × the V3 substrate
+
+> **READ BY:** integration-lead, truth-architect, trajectory-cartographer,
+> convergence-architect, dto-soa-savant, v3-envelope-auditor — and any session
+> designing the OCR→KG→Markov endgame (tesseract `doc.v1` → AriGraph-style KG →
+> GraphRAG retrieval → DeepNSM temporal-Markov context).
+>
+> **Status:** synthesis (2026-07-18). Paper briefs = 6 Opus readers + MDPI
+> fetch (this session). Substrate rows = v3 harvest, source-verified against
+> `canonical_node.rs` / `class_view.rs` / `facet.rs` / `ocr.rs`. Honesty flags
+> [G]/[H]/proposed carried through; nothing silently promoted.
+
+---
+
+## 0 — The probe that lets every representation share ONE SoA
+
+**`PROBE preset-vs-dispatch`** (`.claude/board/EPIPHANIES.md:1184`; reference
+pattern `contract/src/ocr.rs:104-115`). "All tenants can be in every SoA" is the
+**`ValueSchema::Full`** variant — compile-proven to equal *every* tenant
+(`canonical_node.rs:1127`: `assert!(Full.field_mask().count() == VALUE_TENANTS.len())`).
+The write path is a pure function of which tenants the class's schema materialises:
+
+```rust
+let schema = classid_read_mode(classid).value_schema;   // ocr.rs:105
+if schema.has(ValueTenant::EntityType) { write [o..o+2] }   // 108
+if schema.has(ValueTenant::Energy)     { write [o..o+4] }   // 112
+```
+
+One writer, one 512-B row (`key16|edges16|value480`), each lane written **iff the
+live schema carries it**. So the representations below are not separate stores —
+they are **tenant lanes / ClassView readings of one SoA**, selected by classid.
+"One SoA, never transformed" (`EPIPHANIES:5816` §11.1).
+
+## 1 — Inventory of the 7 papers (6 arXiv + MDPI)
+
+| Paper | Native representation | Format | Witness reference | In workspace? |
+|---|---|---|---|---|
+| **AriGraph** 2407.04363 | semantic SPO + **episodic edge** (episode → all its triplets) | triplet + episodic incidence | episode vertex `vₑᵗ=oᵗ` | SPO ✓ · episodic edge **MISSING** (string copy) |
+| **PersonalAI** 2506.17001 | object SPO + **thesis** hyper-edge (semantic) + **episodic** hyper-edge (temporal) | 3 vertex types + 2 hyper-edges | passage node `vᵢₑ=dᵢ` | object ✓ · thesis **MISSING** · episodic hyper-edge **MISSING** (basins() is a derived co-occurrence partition, not episode→triplet membership) · temporal=Lance versions |
+| **Document GraphRAG** MDPI | Document-KG (chapter/section/chunk URIs) + keyword IKG | hierarchical doc tree + keyword edges | chunk URI | doc-tree = vertical radix walk (planned doc-W4) |
+| **SAP Practical** 2507.03226 | dep-parse SPO + **RRF fusion** (vector⊕graph⊕relation) | 3 separate embeddings + rank-fusion | chunk id | SPO ✓ (DeepNSM) · **RRF MISSING** |
+| **StepChain** 2510.02827 | sub-question set + **BFS evidence-chain** `Πsᵤ` | decomposed queries + ordered paths | passage (on-the-fly) | BFS ✓ · **decomposition + path-structure MISSING** |
+| **GraphRAG-FI** 2503.13804 | filtered set + logits-integration tier | two-tier {coarse,fine} + demote | attention proxy | MUL/DK ✓ (VALIDATES) · filter shape adoptable |
+| **GraphRAG under Fire** 2501.14050 | poison relation `r*` (same S+P, diff O) + provenance-trust | relation + per-chunk trust | **source provenance** | source_url ✓ but **trust-wire MISSING** |
+
+## 2 — The facet ladder (le-contract §3) — where each representation addresses
+
+Atom = **16 B** = `4 B prefix (domain|appid|classview-u16) + 96-bit payload`;
+ClassView selects the reading (slot purity — labels never in a payload slot).
+
+| # | Shape | Semantics | L4-exact? |
+|---|---|---|---|
+| L1 | 6×(8:8) | `part_of:is_a` — **episodic basins rail** | — |
+| L2 | 6×(8:8) | `memberof:members` | — |
+| L3 | 6×(8:8) | `mereology:taxonomy` | — |
+| **L4** | 6×(8:8) | **`palette256²`** — each pair indexes 256×256 dist/compose LUT (Fisher-z ρ≥0.999); Morton 2bit×2bit 4×4 perturbation carrier | **YES — similarity = one table read** |
+| L5 | 4×(8:8:8) | SPO-style triplets | — |
+| L6 | 3×(8:8:8:8) | SPOG quads (Odoo) | — |
+| L7 | 2×48-bit | `hhtl ++ helix` — absolute location (q2 FMA) | — |
+| L8 | 2×48-bit | `helix ++ CAM_PQ` — analog old style | — |
+
+Extended 6×(8:8) readings: `area:location` (stacked/**vertical**), `basin:relationtype`
+(**horizontal**), `relationtype:relationtype_orthogonal` (static-basin).
+
+> **Reading A (operator, this session)** — `2×(basin:identity SPO)`: the 6 pairs
+> grouped 3+3 as `[semantic SPO | witness SPO]`, each S/P/O a `(basin:identity)`
+> pair. **This is a NOT-YET-SANCTIONED extended reading** of the L1–L4 plane
+> (distinct from L5's 4×3-byte triplets — it keeps `u8:u8` pairs, grouped
+> semantically). It closes the AriGraph episodic-edge gap in-register. Falsifier:
+> does the `basin` byte agree with the Leiden `communities()`/`basins()`
+> partition, per-typed-plane (never pooled)? [proposed — probe-gated]
+
+## 3 — The 13 value tenants (canonical_node.rs:830-993) + the 14th
+
+| # | Tenant | Width | Carries | witness·context·basin·time·NARS·edge | Built |
+|---|---|---|---|---|---|
+| 0 | **Meta** (MetaWord) | 8B | awareness/NARS/free-energy bits | W·—·—·—·**N**·— | ✅ |
+| 1 | Qualia | 8B | 16×i4 channels | angle | ✅ |
+| 2 | **MaterializedEdges** | 32B | 4× CausalEdge64 | **W·C·—·t·N·E** | ✅ |
+| 3 | Fingerprint | 32B | identity print (points-to) | context(id) | ✅ |
+| 4 | HelixResidue | 6B | 48-bit place (Signed360) | basin/location | ✅ |
+| 5 | TurbovecResidue | 16B | PQ32×4 residue | search | ✅ |
+| 6 | Energy | 4B | spatio-temporal accumulator | time(partial) | ✅ |
+| 7 | Plasticity | 4B | persisted plasticity | learn | ✅ |
+| 8 | EntityType | 2B | OGIT class ordinal | class | ✅ |
+| 9 | **Kanban** | 8B | phase\|exec\|cycle(u32) | **time(cycle)·resolution** | ✅ |
+| 10-12 | **Style triangle** (Frozen/Learned/Explore) | 3×12B | 12 palette256 policy atoms | policy (NOT a trajectory) | ✅ Full-only |
+| 13 | **BoardAggregates** | — | per-mailbox board aggregates | board·time | ❌ **PLANNED** (W2a, offset 188, batched mint) |
+
+`ValueSchema` presets: **Bootstrap**(0) · **Cognitive**(7, 66B) · **Compressed**(4,
+56B) · **Full**(13, 156B → 324 B reserve). All layout-preserving.
+
+## 4 — THE MATRIX (every representation × the questions)
+
+Legend: ✓ yes · ✓✓ *is* this · — no · V/H/E = vertical/horizontal/edge axis.
+
+| Representation | Format | Witness ref | Wit | Ctx | Basin | Axis | Time | NARS | Causality-traj? | Wire |
+|---|---|---|---|---|---|---|---|---|---|---|
+| **SPO triple** | `Triplet{s,p,o,truth}` / spo:u64=3×u8 | WitnessCorpus source_url | ✓ | ✓✓ | ✓ | E | — | ✓ | **✓** owned+wit+NARS | chain episodic_search (Eq.1) |
+| **Episodic-witness edge** | `EpisodicEdges64` 4×u16 / WitnessEntry | episode/passage node | ✓✓ | ✓ | ✓✓ | E+H | ✓ | ~ | **✓** once wired (edge type MISSING as first-class) | wire `EpisodicEdges64` (unwired) |
+| **6×(8:8) facet (Reading A)** | `classid\|2×(basin:id SPO)` | 2nd (witness) SPO | ✓ | ✓ | ✓✓ | V/register | — | — | candidate (static row) | ratify as §3 reading + basin-byte probe |
+| **Community/basin partition** | `Communities` + `EpisodicBasins` | — (derived) | — | ✓ | ✓✓ | H | ~ | — | **grounds** trajectories | + thesis (3rd partition) |
+| **Temporal standing-wave** | `SoaWavePrimer{±radius}`→`WaveProjection`; temporal.rs stream | Lance version / stream pos | ✓ | ✓✓ | ✓ | **T** | ✓✓ | — (leashed) | **✓✓** the WAVE IDENTIFIES | **D-MTS-1** keystone probe (un-run) |
+| **CausalEdge64** | u64: rung\|mantissa\|W-slot\|temporal | W-slot witness arc | ✓✓ | ✓✓ | — | E | ✓ | ✓✓ | **✓✓✓** the trajectory particle | D-MTS-6b shrink; CascadeChannels8 |
+| **Palette / Fisher-z metric** | L4 6×(8:8) palette256² / `PaletteDistanceTable` 256² | — (metric) | — | — | ✓ | L4 addr | — | — | — enables matching | D-MTS-2/3 shader cert |
+| **COCA-4096 constant** | 4096² u8 matrix / cascade L3 / 96D PQ codebook | IS certification witness | ✓ | ✓ | — | L3 rung | — pinned v3 | — | — reference frame | anchor doc (à la Jina-v5 registry) |
+| **Thesis vertex** *(unbuilt)* | community-within-one-episode | episode | ~ | ✓ | ✓✓ | H | — | — | grounds | parallel to `EpisodicBasins` |
+| **RRF fusion** *(unbuilt)* | `rrf(lists,k=60)→Vec<ScoredId>` | — | — | ✓ fuses | ✓ fuses | combinator | — | — | — | free fn in `retrieval.rs` — **keystone** |
+| **Sub-question decomp** *(unbuilt)* | query→{q₁..qₘ} seed sets | — | — | ✓ | — | — | seq | — | LLM/DeepNSM tail | `retrieve_multi` on `OsintRetriever` |
+| **Evidence-chain `Πsᵤ`** *(unbuilt)* | pred-map + `A-(r)→B-(r)→C` | the path (audit trail) | ✓ | ✓✓ | — | E-path | ord | ✓ | **✓✓** chain = trajectory | extend `get_associated`→`associated_paths` |
+| **Provenance→trust** *(unbuilt)* | source_url→`TruthValue.confidence` | source_url | ✓✓ | — | — | — | — | ✓✓ | enables trust-gated traj | at `promote_to_spo`/`WitnessCorpus::insert` |
+| **BeamSearch** *(unbuilt)* | N semantic paths, dedup | — | — | ✓ | — | E-path | — | — | produces path candidates | `retrieve_beam` on `OsintRetriever` |
+| **BoardAggregates** *(unbuilt)* | 14th tenant, offset 188 | — | — | board | — | lane | ✓ | — | — board state | W2a batched mint |
+| **Temporal hyper-edge** | `QueryReference::at(v,rung)` + `Episode.step` | Lance version | ✓ | ✓ | — | **T** | ✓✓ | — | **✓** temporal carrier | shipped (temporal.rs); D-MTS-1 parity |
+
+## 5 — Causality-trajectory candidates (the qualifying subset)
+
+A representation is a **causality-trajectory candidate** iff **owned**
+(`mailbox_owner()`/write-on-behalf) + **witnessed** (a provenance ref) +
+**NARS-revised** (truth freq/conf) + reachable by **proper context edges**
+(`CausalEdge64`). Grounded verdict (v3 harvest §E):
+
+- **✓✓✓ the trajectory itself:** `CausalEdge64`/MaterializedEdges (edge + NARS
+  mantissa + W-slot witness + owner-provenance); the **temporal.rs stream /
+  Lance-versions** (episodic memory = the carrier); the standing-wave **proposes**
+  onto it (leashed to the CE64 particle that confirms — `markov_soa`).
+- **✓✓ trajectory candidates:** SPO triple (with truth), episodic-witness edge
+  (once wired), evidence-chain `Πsᵤ` (the assembled path IS a candidate).
+- **grounds, not is:** Communities/basins (basins GROUND, SPO ANCHOR, the WAVE
+  IDENTIFIES — `E-THINKING-SPINE-CHESS-EVIDENCE-1`); L1–L3 rails; Meta/MetaWord.
+- **not a trajectory:** palette/Fisher-z (metric), COCA (constant frame),
+  Fingerprint/Helix/Turbovec/EntityType/Plasticity (bare values), the style
+  triangle (policy the trajectory dispatches through, explicitly NOT a gestalt).
+
+## 6 — All-tenants-in-one-SoA verdict
+
+Under `ValueSchema::Full` + the `ocr.rs` write-path dispatch, the **tenant-lane**
+representations (in-row) are: SPO (via edges+fingerprint), CausalEdge64
+(MaterializedEdges), the facet register (L1–L4 readings), Kanban (resolution),
+Energy (temporal-partial), the style triangle, and BoardAggregates (once minted).
+The **derived / out-of-row** representations are: Communities/basins/thesis
+(computed partitions), the temporal standing-wave (a projection over the stream of
+rows), COCA/palette (build-time codebooks + LUTs), RRF/decomposition/BeamSearch
+(retrieval logic on `OsintRetriever`). Every reader **re-resolves** the substrate
+from the classid in the 16-B key — no `ClassRoutingDTO`, nothing crosses a mailbox
+boundary (`EPIPHANIES:1184`).
+
+## 7 — Priority wire order (highest ROI first)
+
+1. **RRF fusion** — the keystone; every leg exists, one free fn closes it.
+2. **Provenance→trust** — source_url→confidence at promote; core for OCR mixed-trust docs.
+3. **Chained episodic_search (Eq.1)** — makes the crate actually AriGraph, not RAG.
+4. **Evidence-chain `Πsᵤ`** — surfaces the audit trail the NARS substrate exists for; = causality-trajectory candidates.
+5. **Thesis partition** — the most load-bearing memory type (PersonalAI Table 3).
+6. **Sub-question decomposition** + **BeamSearch** — the two winning retrieval halves.
+7. **D-MTS-1** (Markov-temporal-stream parity) — the probe gating the whole VSA→stream migration.