fix(web): keep manual panel choices during a turn

[t3dotgg]

Turn completion and new PR links could replace a panel the user selected while the turn was running, and could reopen a panel the user hid. resolveProactiveTurnDiffAction only stepped aside for an open pull request, then ChatView called open(activeThreadRef, "diff"), which activates the diff over the Files selection and sets isOpen.

The right panel store now keeps a per-thread count of user panel choices. Every existing action is a user choice by construction. Only openProactive and the two reconcile actions are automatic, so a new action cannot forget to declare itself. ChatView reads the count when it observes a turn or link, and openProactive refuses when the count moved after that. A linked PR still stays above a completed-turn diff. The linked-PR effect and the diff effect are one effect that shares one observation. The count is session-only and is not persisted.

Removed from the original version to reduce risk: the removeSurfaces rewrite of closeSurface, closeAllSurfaces, and closeSurfacesToRight (main's implementations and fallback-tab rule stay), the closeSurfaces batch action, the source argument on terminal close, and the DiffPanel scope change that activated the diff surface. None of these were needed for the fix.

Reworked by Claude Fable 5.1 in Claude Code.

Note

Keep manual right-panel choices during a turn by adding per-thread user-action revisions to rightPanelStore

• Adds a session-only per-thread user-action revision map to the right-panel store; every manual panel action (open, close, toggle, etc.) increments the revision, while automatic updates and resource reconciliation do not
• Introduces rightPanelStore.openProactive, which accepts an expected revision and rejects stale automatic diff or pull-request openings; automatic diffs are also rejected while a pull-request surface is active
• Consolidates the separate turn-diff and pull-request observation effects in ChatView.tsx into one effect using observeProactivePanelUserChoice, which records revisions even during loading and routes both automatic pull-request and diff requests through the revision-checked store method
• Removes the active-surface-kind argument from resolveProactiveTurnDiffAction; the resolver now decides based only on checkpoint and repository state
• Exports the pullRequestSurface constructor so ChatView can build replacement surfaces through the store's canonical path
• Risk: openProactive rejects any automatic request whose expectedRevision does not match the current per-thread revision in rightPanelStore.ts; callers in ChatView.tsx must capture the revision at observation time and pass it through, or proactive panels silently fail to open

^{Macroscope summarized 962cc34.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR changes production right-panel behavior across ChatView and the shared panel store, adding session-scoped revision tracking to coordinate manual choices with turn and pull-request auto-opening. The focused tests reduce risk, but the cross-component state and asynchronous ordering logic is substantial enough to warrant human review.

^{You can add or adjust custom eligibility rules. Learn more.}

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	13.6 KiB	13.6 KiB	+24 B (+0.2%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.0 KiB	7.0 KiB	−6 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.6 KiB	+30 B (+0.4%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	57.1 KiB	57.1 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	10	10	0 (0.0%)	21	✅
Claude	Total thread wire	13.5 KiB	13.6 KiB	+72 B (+0.5%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.0 KiB	7.0 KiB	0 B (0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.5 KiB	6.6 KiB	+72 B (+1.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.8 KiB	57.9 KiB	+88 B (+0.1%)	66.4 KiB	✅
Claude	Live turn messages	8	10	+2 (+25.0%)	21	✅

Baseline: eb8ed80 · PR result: 962cc34 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 113.8 KiB
• Claude decoded thread snapshot: 114.5 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.