test(win): Pro-on-Windows integration + capture/settings fixes

[Anurag-Wednesday]

Windows Pro integration test branch — combines the open Pro-on-Windows PRs (#74 Day+Notifications, #73 LLM lazy settings, #72 Reflect) and adds fixes found while testing the build. Pairs with desktop-pro test/win-pro-integration-0806 (PR #40).

Fixes in this branch

• feat(win): show Capture & Proactive settings on Windows — the Settings screen stayed on the old blanket !isMac gate after the feature nav was migrated to the per-feature seam, so on Windows the Capture health panel (frame/observation counts) and the Proactive-delivery toggle were hidden behind a "Pro on macOS" placeholder even though the engine runs. Now renders the registered section on every Pro platform.
• feat(capture): name a too-small context as the reason observations stop — MIN_OBSERVATION_CTX floor, a pure isContextOverflowError classifier (used pro-side to mark the failure terminal), and a Settings context-window hint that warns when the effective window is too small for observations. Fixes the silent empty-Day/Reflect trap.

Tests

llama-error + ctx-options extended; Settings.pro-sections D31 seam updated to assert Windows Pro renders Capture. Core tsc (node+web) + pro tsc clean; pre-push coverage gate green.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Xr51GbAPDvPYL5yhXj4gva

Summary by CodeRabbit

• New Features

  • Pro Day, Reflect, and Notifications features are now available on supported Windows devices.
  • The app now opens to the most relevant available view based on platform and subscription status.
  • Capture settings are displayed across supported platforms when available.

• Bug Fixes

  • Settings now load reliably after configuration changes and across relaunches.
  • Added clearer warnings when context size is too small for on-device observations.
  • Improved detection of context-limit errors for more helpful failure handling.

Greptile Summary

This PR enables additional Pro surfaces on Windows, lazily loads persisted LLM state after runtime path setup, and adds context-window diagnostics.

• Marks Day, Reflect, and Notifications as Windows-supported and uses the catalog capability seam for the initial route.
• Renders the registered capture settings contribution on Windows Pro.
• Adds context-overflow classification, a minimum observation-context warning, and regression tests.
• Defers active-model and inference-settings reads until first LLM service use.

Confidence Score: 4/5

The PR appears safe to merge, with a non-blocking startup flicker that can misleadingly show Windows Pro users the Day purchase screen before Pro registration finishes.

The newly selected Windows Day landing route renders before the asynchronous Pro route registration, causing the existing null-view fallback to appear temporarily; no blocking runtime or data-integrity failure was established.

Files Needing Attention: src/renderer/src/App.tsx

Important Files Changed

Filename	Overview
src/main/llm.ts	Defers persisted model and inference-settings loading until first use; checked entry points load state before consuming persisted fields.
src/main/llama-error.ts	Adds a pure context-overflow classifier, though no production caller exists in this repository.
src/renderer/src/App.tsx	Uses the feature catalog for the initial route, which introduces a transient UpgradeScreen fallback while Windows Pro registration is pending.
src/renderer/src/components/Settings.tsx	Removes the macOS-only gate from the registered capture settings contribution.
src/renderer/src/components/pro/proCatalog.ts	Declares Day, Reflect, and Notifications supported on Windows and centralizes initial-view selection.
src/renderer/src/lib/ctx-options.ts	Warns when the effective context is below the shared observation-processing threshold.
src/shared/llm-defaults.ts	Defines the shared 4096-token minimum used for observation-context diagnostics.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart LR
  A[Windows Pro launch] --> B[landingView selects Day]
  B --> C{Pro renderer registered?}
  C -- No --> D[UpgradeScreen fallback]
  E[Async Pro activation] --> F[Register Pro renderer]
  F --> C
  C -- Yes --> G[Render Day]

Loading

_{Reviews (1): Last reviewed commit: "test(win): assert Windows Pro renders th..." | Re-trigger Greptile}

Greptile also left 1 inline comment on this PR.

[qodo-code-review]

ⓘ Qodo reviews are paused because your trial has ended. Ask your workspace admin to add credits to resume reviews. Manage billing

[coderabbitai]

📝 Walkthrough

Walkthrough

The PR adds lazy persisted-settings loading, context-window overflow detection and warnings, Windows support for selected Pro features, platform-aware landing navigation, and cross-platform capture settings rendering.

Changes

LLM context and initialization

Layer / File(s)	Summary
Context-size contract and warnings src/shared/llm-defaults.ts, src/renderer/src/lib/ctx-options.ts, src/renderer/src/lib/__tests__/ctx-options.test.ts	Adds MIN_OBSERVATION_CTX and warns when the effective context window is below 4K.
Context overflow detection src/main/llama-error.ts, src/main/__tests__/llama-error.test.ts	Adds isContextOverflowError and tests for recognized and excluded llama-server messages.
Lazy LLM state loading src/main/llm.ts, src/main/__tests__/llm-lazy-settings-load.test.ts	Defers model and persisted-settings loading until first use and initializes state before public operations.

Windows Pro catalog and settings

Layer / File(s)	Summary
Platform-aware Pro catalog src/renderer/src/components/pro/proCatalog.ts, src/renderer/src/lib/__tests__/proCatalog.lookup.test.ts, src/renderer/src/App.tsx	Adds Windows support for Day, Reflect, and Notifications. Landing navigation now uses landingView.
Cross-platform capture settings src/renderer/src/components/pro/proSettingsCatalog.ts, src/renderer/src/components/Settings.tsx, src/renderer/src/components/__tests__/Settings.pro-sections.test.tsx	Removes Mac-only capture and proactive gating. Windows Pro settings now render registered capture content.

Estimated code review effort: 4 (Complex) | ~45 minutes

Possibly related PRs

• off-grid-ai/OGAD#46: Overlaps with the LLMService and context-overflow handling changes.
• off-grid-ai/OGAD#49: Overlaps with Windows feature support and proCatalog.ts tests.
• off-grid-ai/OGAD#73: Overlaps with lazy LLMService settings loading and regression tests.

Suggested reviewers: alichherawalla

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 75.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly summarizes the main Windows Pro integration and capture/settings changes in the pull request.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

✨ Finishing Touches 💡 1

📝 Generate docstrings 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch test/win-pro-integration-0806

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[greptile-apps]

Day landing flashes the upsell

On Windows Pro launches, landingView selects Day before the asynchronous Pro renderer is registered, so the null-view fallback briefly displays the Day upgrade screen to an entitled user; slower activation leaves the misleading upsell visible longer.

[coderabbitai]

Actionable comments posted: 3

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Inline comments:
In `@src/main/__tests__/llm-lazy-settings-load.test.ts`:
- Around line 47-57: Add a regression test alongside the existing lazy settings
case that seeds a model file and active-model.json, constructs LLMService before
configureRuntime(), then configures the profile and verifies activeModelInfo()
resolves the model from that profile on first use. Ensure the test specifically
covers deferred active-model resolution through ensureLoaded().

In `@src/renderer/src/App.tsx`:
- Around line 229-232: Update the root-path routing effect in App.tsx to use
landingView(currentPlatform(), isPro) instead of unconditionally selecting day,
preserving the selector’s capability-based result for free, Windows Pro, and
unsupported-platform users. Add App-level root-path integration coverage in
src/renderer/src/lib/__tests__/proCatalog.lookup.test.ts lines 173-212 for those
three user scenarios; both sites require changes.

In `@src/renderer/src/lib/ctx-options.ts`:
- Around line 42-48: The effective-context check in ctx-options.ts#L42-L48 must
run before the model-cap hint so it takes priority when the engine’s effective
context is below MIN_OBSERVATION_CTX; preserve the existing warning text and
other hint behavior. Add a regression assertion in
src/renderer/src/lib/__tests__/ctx-options.test.ts#L62-L73 using ctxSize 65536,
effectiveCtxSize 2048, and modelMaxCtx 32768, verifying the observation warning
is returned instead of the model-cap hint.

🪄 Autofix

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro Plus

Run ID: 6155446d-2795-42ad-876a-e0a5e484adb8

📥 Commits

Reviewing files that changed from the base of the PR and between efcb0e9 and 7c528a0.

📒 Files selected for processing (13)

• src/main/__tests__/llama-error.test.ts
• src/main/__tests__/llm-lazy-settings-load.test.ts
• src/main/llama-error.ts
• src/main/llm.ts
• src/renderer/src/App.tsx
• src/renderer/src/components/Settings.tsx
• src/renderer/src/components/__tests__/Settings.pro-sections.test.tsx
• src/renderer/src/components/pro/proCatalog.ts
• src/renderer/src/components/pro/proSettingsCatalog.ts
• src/renderer/src/lib/__tests__/ctx-options.test.ts
• src/renderer/src/lib/__tests__/proCatalog.lookup.test.ts
• src/renderer/src/lib/ctx-options.ts
• src/shared/llm-defaults.ts

[coderabbitai]

📐 Maintainability & Code Quality | 🟡 Minor | ⚡ Quick win

Add a lazy active-model regression case.

These tests cover llm-settings.json only. ensureLoaded() also defers active-model.json resolution.

Seed a model file and active-model.json, construct LLMService before configureRuntime(), then assert that activeModelInfo() reads the model from the profile configured at first use.

As per coding guidelines, "Add regression tests in the same change for every behavior change, including bug cases, branches, conditions, error paths, and copy or contract changes."

Proposed regression test

+const seedActiveModel = (dataDir: string, primary: string): void => {
+  const modelsDir = path.join(dataDir, 'models')
+  fs.mkdirSync(modelsDir, { recursive: true })
+  fs.writeFileSync(path.join(modelsDir, primary), '')
+  fs.writeFileSync(
+    path.join(modelsDir, 'active-model.json'),
+    JSON.stringify({ id: 'late-model', primary })
+  )
+}
+
+it('reads the active model from the profile configured at first use', () => {
+  const svc = new LLMService()
+  seedActiveModel(tmp, 'late-model.gguf')
+  configureRuntime({ dataDir: tmp })
+
+  expect(svc.activeModelInfo()).toEqual({ id: 'late-model', vision: false })
+})

📝 Committable suggestion

‼️ IMPORTANT
Carefully review the code before committing. Ensure that it accurately replaces the highlighted code, contains no missing lines, and has no issues with indentation. Thoroughly test & benchmark the code to ensure it meets the requirements.

Suggested change

	it('picks up a data dir configured AFTER the instance was constructed', () => {
	// Construct FIRST — mirrors the module-level singleton being built during imports.
	const svc = new LLMService()
	// ...then point the runtime at the profile, as index.ts's body does later.
	seedSettings(tmp, { performanceMode: 'extreme', temperature: 0.42 })
	configureRuntime({ dataDir: tmp })
	const s = svc.getSettings()
	expect(s.performanceMode).toBe('extreme')
	expect(s.temperature).toBe(0.42)
	})
	const seedActiveModel = (dataDir: string, primary: string): void => {
	const modelsDir = path.join(dataDir, 'models')
	fs.mkdirSync(modelsDir, { recursive: true })
	fs.writeFileSync(path.join(modelsDir, primary), '')
	fs.writeFileSync(
	path.join(modelsDir, 'active-model.json'),
	JSON.stringify({ id: 'late-model', primary })
	)
	}
	it('reads the active model from the profile configured at first use', () => {
	const svc = new LLMService()
	seedActiveModel(tmp, 'late-model.gguf')
	configureRuntime({ dataDir: tmp })
	expect(svc.activeModelInfo()).toEqual({ id: 'late-model', vision: false })
	})

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/main/__tests__/llm-lazy-settings-load.test.ts` around lines 47 - 57, Add
a regression test alongside the existing lazy settings case that seeds a model
file and active-model.json, constructs LLMService before configureRuntime(),
then configures the profile and verifies activeModelInfo() resolves the model
from that profile on first use. Ensure the test specifically covers deferred
active-model resolution through ensureLoaded().

Source: Coding guidelines

[coderabbitai]

🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

Keep root-path routing consistent with landingView.

landingView returns 'models' for free users and unsupported platforms. The mount-only URL effect still maps '/' to 'day' at Line 295. That effect runs after Line 232 and overwrites the selector result.

• src/renderer/src/App.tsx#L229-L232: route / through landingView(currentPlatform(), isPro) instead of the unconditional day mapping.
• src/renderer/src/lib/__tests__/proCatalog.lookup.test.ts#L173-L212: add an App-level root-path test for free, Windows Pro, and unsupported-platform users. The current selector-only tests cannot detect this overwrite.

As per coding guidelines, "Define mappings, routing rules, and capability checks once and reuse the single source of truth" and "Add user-behavior integration tests through real product boundaries."

📍 Affects 2 files

• src/renderer/src/App.tsx#L229-L232 (this comment)
• src/renderer/src/lib/__tests__/proCatalog.lookup.test.ts#L173-L212

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/renderer/src/App.tsx` around lines 229 - 232, Update the root-path
routing effect in App.tsx to use landingView(currentPlatform(), isPro) instead
of unconditionally selecting day, preserving the selector’s capability-based
result for free, Windows Pro, and unsupported-platform users. Add App-level
root-path integration coverage in
src/renderer/src/lib/__tests__/proCatalog.lookup.test.ts lines 173-212 for those
three user scenarios; both sites require changes.

Source: Coding guidelines

[coderabbitai]

🎯 Functional Correctness | 🟠 Major | ⚡ Quick win

Prioritize the effective-context warning before the model-cap hint.

When ctxSize exceeds modelMaxCtx and effectiveCtxSize is below 4K, Line 39 returns before this branch. The UI then hides the warning for the context size that the engine actually uses.

Move the effective-context check before the model-cap return. Add a regression case with ctxSize: 65536, effectiveCtxSize: 2048, and modelMaxCtx: 32768.

• src/renderer/src/lib/ctx-options.ts#L42-L48: evaluate the below-floor effective context before existing hint branches.
• src/renderer/src/lib/__tests__/ctx-options.test.ts#L62-L73: assert that the observation warning wins when the model-cap condition is also true.

📍 Affects 2 files

• src/renderer/src/lib/ctx-options.ts#L42-L48 (this comment)
• src/renderer/src/lib/__tests__/ctx-options.test.ts#L62-L73

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@src/renderer/src/lib/ctx-options.ts` around lines 42 - 48, The
effective-context check in ctx-options.ts#L42-L48 must run before the model-cap
hint so it takes priority when the engine’s effective context is below
MIN_OBSERVATION_CTX; preserve the existing warning text and other hint behavior.
Add a regression assertion in
src/renderer/src/lib/__tests__/ctx-options.test.ts#L62-L73 using ctxSize 65536,
effectiveCtxSize 2048, and modelMaxCtx 32768, verifying the observation warning
is returned instead of the model-cap hint.

Source: Coding guidelines

[sonarqubecloud]

Quality Gate passed

Issues
2 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud