fix(server): stop routine events from rescanning thread history

[t3dotgg]

Assistant messages and routine tool activity reload complete thread histories even when no thread summary can change. This slows both the default buffered mode and legacy streaming, and busy threads can delay other conversations.

Skip the summary refresh for assistant messages and routine activities. Keep existing refreshes for user messages, approvals, user input, plans, session changes, and reverts.

Thanks to @cheruvian for #5855, @dain for #7356, and @maria-rcks for #7486.

Fixes #4008.

Checks:

• vp test run apps/server/src/orchestration/Layers/ProjectionPipeline.test.ts
• vp run --filter t3 typecheck
• vp check apps/server/src/orchestration/Layers/ProjectionPipeline.ts apps/server/src/orchestration/Layers/ProjectionPipeline.test.ts

Model and harness: GPT-5.6 Sol in Codex.

---

Note

Medium Risk
Changes when derived projection_threads summary columns update; anything expecting fresh counts after assistant or routine tool activity may stay stale until a user message or allowlisted event.

Overview
Stops expensive thread shell summary refreshes on events that cannot change sidebar/summary fields (latestUserMessageAt, pending approval/input counts, actionable plans).

Adds shouldRefreshThreadShellSummary and uses it before refreshThreadShellSummary in the shared message/activity/plan handler. Assistant thread.message-sent events and most thread.activity-appended kinds (e.g. tool.updated) no longer trigger a full history rescan; user messages and allowlisted approval/user-input activity kinds still do. Session, revert, and proposed-plan paths are unchanged.

The projection pipeline test uses a projection_threads update trigger to assert assistant messages and routine tool activity each produce only the lightweight updatedAt upsert—not an extra summary refresh.

^{Reviewed by Cursor Bugbot for commit ec7c00a. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Guard refreshThreadShellSummary call in applyProjectsProjection to skip routine events

Introduces shouldRefreshThreadShellSummary to decide whether a given event warrants a thread shell summary refresh. It returns true for user messages and a specific set of activity kinds (approval.requested, approval.resolved, provider.approval.respond.failed, user-input.requested, user-input.resolved, provider.user-input.respond.failed), and false for assistant messages and other activity kinds.

• Replaces the unconditional refreshThreadShellSummary(threadId) call in ProjectionPipeline.ts with a conditional call guarded by this function. The thread row upsert with updatedAt still runs on every event.
• Adds a test in ProjectionPipeline.test.ts using a SQL trigger to count projection_threads updates, verifying that a user message triggers one refresh and a subsequent tool.updated activity does not trigger an additional one.

^{Macroscope summarized ec7c00a.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 8db5e17a-95a7-445e-b284-86a46aab3afa

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

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.}

[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	13.4 KiB	13.3 KiB	−100 B (−0.7%)	15.1 KiB	✅
Codex	Thread snapshot wire	6.9 KiB	6.9 KiB	−5 B (−0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.5 KiB	6.5 KiB	−95 B (−1.4%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	55.0 KiB	55.6 KiB	+602 B (+1.1%)	66.4 KiB	✅
Codex	Live turn messages	16	11	−5 (−31.3%)	21	✅
Claude	Total thread wire	13.4 KiB	13.3 KiB	−105 B (−0.8%)	15.1 KiB	✅
Claude	Thread snapshot wire	6.9 KiB	6.9 KiB	+6 B (+0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.6 KiB	6.4 KiB	−111 B (−1.7%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	55.8 KiB	56.4 KiB	+620 B (+1.1%)	66.4 KiB	✅
Claude	Live turn messages	16	11	−5 (−31.3%)	21	✅

Baseline: 9996038 · PR result: ec7c00a · 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: 109.4 KiB
• Claude decoded thread snapshot: 110.1 KiB

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

[macroscopeapp]

Approvability

A prior Macroscope approval was dismissed. Re-evaluating ec7c00a…

Dismissing prior approval to re-evaluate ec7c00a