🤖 perf: automate workspace-open perf + React render profiling#2397
Merged
🤖 perf: automate workspace-open perf + React render profiling#2397
Conversation
Collaborator
Author
|
Perf follow-up ideas from the latest automated workspace-open profiles (prioritized for "instant open" UX):
I can implement items 1–3 first in a focused follow-up to get measurable improvement with low risk. |
Add automated CDP + React render profiling for workspace-open e2e perf scenarios. - add deterministic history profile fixtures (small/medium/large/tool-heavy/reasoning-heavy) - add reusable Chrome trace/CPU/metrics/heap artifact capture helpers - add perf workspace-open scenario with artifact output + assertions - add React render-path profiling for chat-pane/header/transcript/input - add Makefile target and scheduled/manual workflow for perf collection --- _Generated with `mux` • Model: `openai:gpt-5.3-codex` • Thinking: `xhigh` • Cost: `$0.00`_ <!-- mux-attribution: model=openai:gpt-5.3-codex thinking=xhigh costs=0.00 -->
ammar-agent
force-pushed
the
perf-testing-d6tn
branch
from
dcfde52 to
e8eca58
Compare
ammario
approved these changes
Feb 13, 2026
ammario
pushed a commit
that referenced
this pull request
Feb 14, 2026
## Summary Follow-up perf pass for workspace-open rendering that reduces critical-path work in the transcript UI and message utilities. This improves startup paint times for loaded workspaces, especially on long tool-heavy histories. ## Background PR #2397 added automated perf profiling for workspace-open flows. Those artifacts showed recurring renderer-side hot spots during initial transcript paint (message-row prop churn, per-row grouping scans, and unnecessary live subscriptions after completion). ## Implementation - **ChatPane render-path reductions** - Precompute `bash_output` grouping once per message snapshot (`computeBashOutputGroupInfos`) instead of per-row scans. - Pass `taskReportLinking` only to `task` / `task_await` rows. - Precompute stable `userMessageNavigation` objects by `historyId` so non-message state bumps stop invalidating row props. - **Workspace caught-up sequencing** - Defer the caught-up `usageStore.bump()` to idle (`requestIdleCallback` with timeout fallback) so initial transcript paint is prioritized. - **Reasoning row rendering** - Use a plain truncated summary line in collapsed headers. - Avoid rendering full collapsed markdown body unless expanded/streaming. - **Bash row subscriptions** - Gate live-output / latest-streaming / foreground-id subscriptions to rows that still need live state (executing or completed-without-final-output). - **Transcript truncation policy** - Reduce `MAX_DISPLAYED_MESSAGES` from 128 → 64. - Preserve user prompts + structural markers by default; allow older assistant/tool/reasoning rows to stay behind `history-hidden` until “Load all”. - **Desktop dist/e2e stability fix** - Replace `@/...` alias imports in `src/desktop/main.ts` with relative imports so dist runs do not emit unresolved alias `require()` calls. ## Validation - `bun test src/browser/utils/messages/messageUtils.test.ts src/browser/utils/messages/StreamingMessageAggregator.test.ts src/browser/stores/WorkspaceStore.test.ts` - `make static-check` - Perf scenario (dist build path): - `MUX_E2E_RUN_PERF=1 MUX_PROFILE_REACT=1 MUX_E2E_LOAD_DIST=1 xvfb-run -a bun x playwright test tests/e2e/scenarios/perf.workspaceOpen.spec.ts --project=electron --workers=1` ### Perf wall-time (isolated profile runs) | Profile | Before (ms) | After (ms) | Delta | |---|---:|---:|---:| | small | 211 | 202 | -4.3% | | medium | 212 | 186 | -12.3% | | large | 244 | 200 | -18.0% | | tool-heavy | 252 | 185 | -26.6% | | reasoning-heavy | 174 | 171 | -1.7% | ## Risks - **Transcript visibility tradeoff (intentional):** older assistant rows can now be omitted behind `history-hidden` by default; users can restore with “Load all”. - **Reasoning header formatting tradeoff (intentional):** collapsed summary is plain text (no markdown formatting) to avoid markdown render cost in headers. - **Bash subscription gating:** completed rows now skip live subscriptions unless needed; behavior still preserves executing and completed-without-output cases. --- _Generated with `mux` • Model: `openai:gpt-5.3-codex` • Thinking: `xhigh` • Cost: `$4.95`_ <!-- mux-attribution: model=openai:gpt-5.3-codex thinking=xhigh costs=4.95 -->
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Automate workspace-open performance profiling in e2e so runs produce repeatable Chrome/Electron and React artifacts without manual DevTools interaction.
Background
We needed repeatable perf coverage for large-history workspace load flows, plus machine-readable artifacts that can run in CI/nightly and be compared over time.
Implementation
tests/e2e/utils/perfProfile.ts) that captures:perf-summary.jsontests/e2e/utils/historyFixture.ts):small,medium,large,tool-heavy,reasoning-heavytests/e2e/scenarios/perf.workspaceOpen.spec.ts) that:artifacts/perf/**src/browser/utils/perf/reactProfileCollector.ts) and preload/browser flag plumbing (MUX_PROFILE_REACT) so e2e can collect React render data.chat-panechat-pane.headerchat-pane.transcriptchat-pane.inputmake test-e2e-perf.github/workflows/perf-profiles.ymlValidation
make fmtmake typecheckmake lintmake static-checkMUX_E2E_PERF_PROFILES=small xvfb-run -a make test-e2e-perfRisks
Low-to-moderate. Changes are mostly test/perf infrastructure. Runtime impact is gated behind
window.api.enableReactPerfProfile/MUX_PROFILE_REACT, so normal user flows should remain unaffected.Generated with
mux• Model:openai:gpt-5.3-codex• Thinking:xhigh• Cost:$0.00