feat(web): pin turn diff baselines

[sarjann]

What Changed

Add a pinned baseline to the web/desktop diff scope menu, using the server's existing checkpoint-range API.

• Pin baseline compares the saved state after a completed turn with Latest turn. Start of thread (Turn 0) includes the first captured turn's changes.
• Selecting a numbered Turn fixes the comparison endpoint; Latest turn resumes following completed turns. New completed turns preserve the baseline and any fixed endpoint.
• The scope label shows both endpoints. Unpin baseline returns to a single-turn diff; working-tree and branch scopes clear the pin.
• Completed turns with no individual file changes are muted and labeled No changes. They remain selectable when valid for the range; endpoints before the baseline are disabled.

Pins are stored per client, thread, and environment. Comparison caches and review state distinguish both endpoints. Includes focused store/range tests and a short user-guide section.

Why

Work often spans an implementation turn and a small follow-up correction. Reviewing those turns separately makes it harder to see the combined result. Pinning a saved baseline lets users review the net changes across that chunk of work.

Scope is limited to web and the shared desktop UI. No server, contract, mobile, or development-infrastructure changes. Imported CLI sessions can only compare snapshots captured by T3; Turn 0 does not reconstruct earlier CLI history.

UI Changes

Before	After

Pinned comparison labels

Follow latest: Turn 18 → Latest turn keeps Turn 18 as the baseline while the endpoint advances with completed turns. The menu below shows numbered endpoints and the new No changes labels.

Fixed endpoint: choosing Turn N changes the label to Turn 18 → Turn N and keeps both endpoints fixed. This example uses Turn 14 → Turn 16 to show a range with code changes.

Validation

• 120 focused tests passed: diffPanelStore.test.ts and ChatView.logic.test.ts.
• Web TypeScript check passed.
• Targeted lint completed with existing warnings and no errors; formatting and diff whitespace checks passed.
• Manually verified in the current web build: pin Turn 14 to Latest, switch to Turn 16, and unpin back to the single-turn view. Checked Turn 0 availability, disabled earlier endpoints, and No changes labels.
• Mobile is intentionally outside this PR and was not tested.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Implemented with GPT-6-Astra using Codex CLI in T3 Code.

Note

Add pinned baseline support to diff panel turn comparisons

• Adds the ability to pin a completed turn or the thread-start snapshot as a diff baseline, then compare a fixed or latest-following endpoint turn against it via the scope menu in DiffPanel.tsx
• Extends diffPanelStore.ts with DiffBaseline and DiffPanelSelection types, pinBaseline and selectLatestTurn actions, and a getCheckpointDiffRange helper that returns null for unavailable or reversed endpoints
• Updates the completion effect in ChatView.tsx so a pinned-baseline comparison is not replaced when a new turn completes; unpinned selections still auto-select the latest turn
• Reconciliation now independently validates the endpoint and the pinned baseline, clearing invalid baselines and falling back to branch scope when no checkpoints remain
• Adds user-facing documentation for baseline pinning, thread-start comparisons, and scope menu behavior in source-control.md
• Behavioral Change: reconcileTurnSelection in diffPanelStore.ts now clears an unavailable pinned baseline and falls back to branch scope when the available turn list is empty; previously the selection stayed on the removed turn

^{Macroscope summarized e2d94ea.}

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-05T15:46:52.186701Z	00f8001	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 00f8001de0

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Disable Turn 0 when the baseline ref is absent

When Git is initialized during a running turn, CheckpointReactor intentionally records the completion checkpoint without creating the missing turn-0 baseline (CheckpointReactor.ts:240-274; its test confirms the ref remains absent at lines 1285-1287). This item is nevertheless enabled whenever any checkpoint summary exists; selecting it calls the full-thread diff, which always uses checkpointRefForThreadTurn(threadId, 0) with fallback disabled (CheckpointDiffQuery.ts:257-263), so this advertised comparison returns an error instead of a diff. Gate the option on actual turn-0 baseline availability or make the API handle the missing initial snapshot.

Useful? React with 👍 / 👎.

[macroscopeapp]

🟡 Medium components/DiffPanel.tsx:220

Pinned comparisons produce the wrong result for imported summaries without a stored checkpointTurnCount: baselineTurnCount becomes undefined, so getCheckpointDiffRange falls back to the preceding-turn diff and the label/cache omit the pinned baseline. The Turn menu also enables older endpoints with an inferred count because undefined < baselineTurnCount is false, and selecting one then yields no patch. Use inferredCheckpointTurnCountByTurnId[turnId] as the fallback both when computing baselineTurnCount and when disabling menu entries.

🤖 Copy this AI Prompt to have your agent fix this:

In file @apps/web/src/components/DiffPanel.tsx around line 220:

Pinned comparisons produce the wrong result for imported summaries without a stored `checkpointTurnCount`: `baselineTurnCount` becomes `undefined`, so `getCheckpointDiffRange` falls back to the preceding-turn diff and the label/cache omit the pinned baseline. The Turn menu also enables older endpoints with an inferred count because `undefined < baselineTurnCount` is false, and selecting one then yields no patch. Use `inferredCheckpointTurnCountByTurnId[turnId]` as the fallback both when computing `baselineTurnCount` and when disabling menu entries.

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds a persisted, user-facing baseline-comparison workflow across the web diff UI, selection store, and proactive diff opening, rather than a small mechanical change. Unresolved findings cover unavailable Turn 0 baselines and imported summaries, so the behavior warrants human review.

Not approved because:

• 1 blocking correctness issue found at or above your repo's Minimum Blocking Severity

^{Adjust the Minimum Blocking Severity for this repo — including turning it Off — in Settings. You can add or adjust custom eligibility rules. Learn more.}