perf(web): skip checkpoint map rebuilds while streaming

[t3dotgg]

ChatView rebuilt rewind targets on every streaming update. Timeline row reuse depended on callers keeping two Maps stable.

The timeline now accepts checkpoint summaries and owns both lookups. Text-only updates reuse rows before checkpoint maps are rebuilt. Equal control snapshots preserve reuse. Rewind callbacks reject clicks from an old thread or checkpoint snapshot.

Verification

• 301 focused timeline, ChatView logic, and Markdown tests pass after rebasing onto current main.
• A long-thread fixture with 250 historical checkpoints uses the real client reducer, detail selectors, and attachment preview code. Ten text updates do not read checkpoint lookup fields in the timeline projection and keep historical row references. Image URL renewal and turn completion still update the rows.
• A mounted timeline keeps code wrap state through streaming and highlighter recovery. Changed checkpoints send the current rewind target.
• Web typecheck and targeted lint pass. Existing React lint warnings remain.

This changes the shared web and desktop timeline. It does not change mobile, provider behavior, or wire contracts. These are source and component checks, not browser CPU or frame measurements.

In a disposable browser, five streamed text updates kept the same connected code node, line-wrap state, and selected text. Changing the checkpoint summary changed the revert confirmation from checkpoint 1 to checkpoint 3. Both dialogs were cancelled. The same behavior passed on main and the combined integration build.

Browser evidence

Before is main 0c200c5f. After is combined integration 8316ed24. These checks use browser-only RPC events. No provider ran and no checkpoint restore was submitted. This is client behavior evidence, not a CPU or frame-rate measurement.

Before	After

Before	After

Before recording · After recording

Note

Skip checkpoint map rebuilds during streaming by moving revert logic into MessagesTimeline

• Moves buildRevertTurnCountByUserMessageId from ChatView.logic.ts into MessagesTimeline.logic.ts, so checkpoint and revert-count calculation happens inside timeline row derivation instead of in the parent component.
• Changes deriveMessagesTimelineRows input contract from two caller-precomputed Maps to a checkpoint summary array (turnDiffSummaries) and a supportsConversationRollback flag. The row builder now indexes assistant summaries and infers checkpoint counts locally.
• Expands the unchanged-context check in replaceStreamingMessageRows to compare checkpoint-summary arrays, latest-turn values, and expanded-turn/work-group sets, so streaming updates reuse the previous row projection when row-affecting inputs are shallowly equal.
• Replaces the message-ID revert callback with a numeric target-turn-count callback across MessagesTimeline.tsx, UserTimelineRow, and RevertUserMessageButton.
• Risk: deriveMessagesTimelineRows now requires shallow-equal checkpoint-summary arrays to reuse streaming projections; callers passing new array references per render will rebuild rows every update and lose the streaming-skip optimization.

📊 Macroscope summarized 3b43359. 3 files reviewed, 1 issue evaluated, 0 issues filtered, 1 comment posted

🗂️ Filtered Issues

Model

Created with GPT-6 Astra in Codex. Follow-up test refinement and rebase with Codex. Browser verification and evidence upload in Codex.

[cursor]

Bugbot is paused — on-demand spend limit reached

Bugbot uses usage-based billing for this team and has hit its on-demand spend limit.

A team admin can raise the spend limit in the Cursor dashboard, or wait for the next billing cycle to continue.

[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	13.5 KiB	13.5 KiB	+15 B (+0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.0 KiB	7.0 KiB	−3 B (−0.0%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	+18 B (+0.3%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Codex	Live turn messages	8	8	0 (0.0%)	21	✅
Claude	Total thread wire	13.6 KiB	13.6 KiB	−6 B (−0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.0 KiB	7.0 KiB	−1 B (−0.0%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.5 KiB	−5 B (−0.1%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.9 KiB	57.9 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	10	10	0 (0.0%)	21	✅

Baseline: 7544d3d · PR result: 3b43359 · 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.

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This production refactor changes streaming row reuse and passes a row-captured checkpoint count into an irreversible revert operation. A stale row may therefore target the wrong checkpoint after state changes, so the unresolved high-severity risk requires 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.}

Dismissing prior approval to re-evaluate 327c9dd

Dismissing prior approval to re-evaluate 3b43359