The convert step was using find which picked up a 0-byte file from
playwright's videos/ directory instead of the valid qa-session.webm.
Now prefers qa-session.webm explicitly and skips empty files.
Codex was using pnpm dlx instead of the global playwright-cli.
Pre-install chromium in setup step and make prompt explicit about
using the global command directly without pnpm/npx.
Replace claude --print with codex exec for cheaper QA runs.
Uses codex-mini-latest model ($1.50/$6 vs Sonnet $3/$15).
Uses existing OPENAI_API_KEY secret (no new secrets needed).
- Replace saveVideo config (didn't produce video) with explicit
playwright-cli video-start/video-stop commands in QA prompt
- Remove apt-get install ffmpeg step (pre-installed on GH runners)
- Switch video review model from gpt-4o to gpt-4.1-mini
- Enable saveVideo in playwright-cli config for real video recording
- Replace screenshot stitching with webm→mp4 conversion
- Move video review step before deploy so reports are included
- Add GPT video review reports inline on the Cloudflare Pages site
- Each video card now has expandable "GPT Video Review" section
- Set .playwright/cli.config.json with outputDir pointing to screenshots/
- This way bare 'playwright-cli screenshot' auto-saves to the right place
- Create screenshot directory before Claude runs (don't rely on Claude)
- Collect step now searches working directory for stray PNGs
- Simplified prompt: no --filename needed, just 'playwright-cli screenshot'
Screenshots were saved to artifact root but stitch looked in frames/.
Now: prompt tells Claude to save to screenshots/ dir with numbered names,
collect step consolidates PNGs there, stitch step globs from screenshots/.
Removed video-start/video-stop (Claude doesn't use them).
- Add playwright-cli config with outputDir and saveVideo
- Use video-start/video-stop instead of relying on screenshot frames
- Add fallback artifact collection from .playwright-cli/ default dir
- Simplify prompts to focus on video recording workflow
The escaped \$QA_ARTIFACTS in the heredoc produced literal text
'$QA_ARTIFACTS' in the prompt. Claude's Bash tool didn't reliably
expand this env var, so no screenshots or reports were saved.
Remove the escapes so the heredoc expands the variable to the actual
path (e.g. /home/runner/work/_temp/qa-artifacts).
Backtick-wrapped playwright-cli examples in the unquoted heredoc were
being interpreted as bash command substitution, producing empty prompts.
Replace backtick syntax with plain "Run:" prefixed commands.
- Remove all Xvfb/ffmpeg screen recording infrastructure from qa job
(captured blank display since playwright-cli runs headless)
- Add screenshot instructions to QA prompts: Claude saves sequential
frames to $QA_ARTIFACTS/frames/ after every interaction
- Stitch screenshots into video via ffmpeg in report job (2fps)
- Merge video-review job into report job (4 jobs → 3 jobs)
- Unified PR comment with video links + video review in <details> collapse
- Clean up stale QA_VIDEO_REVIEW_COMMENT markers from prior runs
Move extra_server_params input to env var to prevent shell injection
from untrusted input. Replace wait-for-it pip dependency with a
cross-platform curl polling loop.
Add Claude Code skills and a label-triggered QA workflow:
- .claude/skills/comfy-qa/SKILL.md: 12-category QA test plan using
playwright-cli for browser automation
- .github/workflows/pr-qa.yaml: CI workflow triggered by qa-changes
(focused, Linux) or qa-full (3-OS matrix) labels. Records screen via
ffmpeg, runs Claude CLI with playwright-cli, deploys video gallery to
Cloudflare Pages, posts PR comment with GIF thumbnails, and runs
OpenAI vision-based video review
- scripts/qa-video-review.ts: frame extraction + GPT-4o analysis
- scripts/qa-video-review.test.ts: unit tests for video review
- knip.config.ts: resolve knip errors for ingest-types package
## What
- Add `include: ['src/**/*.{ts,vue}']` to vitest coverage config so ALL
source files appear in reports (previously only imported files showed
up)
- Add `lcov` reporter for CI integration and VS Code coverage gutter
- Add `exclude` patterns for test files, locales, litegraph, assets,
declarations, stories
- Add `test:coverage` npm script
## Why
Coverage reports currently only show files that are imported during test
runs. Adding the `include` pattern reveals the true gap — files with
zero coverage that were previously invisible. The lcov reporter enables
IDE integration and future CI coverage comments (Codecov/Coveralls).
## Testing
`npx tsc --noEmit` passes. No behavioral changes — this only affects
coverage reporting configuration.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10575-config-add-vitest-coverage-include-pattern-lcov-reporter-32f6d73d365081c8b59ad2316dd2b198)
by [Unito](https://www.unito.io)
## Summary
Add deterministic mock data fixtures for browser tests so they can use
`page.route()` to intercept API calls without depending on a live
backend.
## Changes
- **`browser_tests/fixtures/data/nodeDefinitions.ts`** — Mock
`ComfyNodeDef` objects for KSampler, CheckpointLoaderSimple, and
CLIPTextEncode
- **`browser_tests/fixtures/data/systemStats.ts`** — Mock `SystemStats`
with realistic RTX 4090 GPU info
- **`browser_tests/fixtures/data/README.md`** — Usage guide for
`page.route()` interception
All fixtures are typed against the Zod schemas in `src/schemas/` and
pass `pnpm typecheck:browser`.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10662-test-add-mock-data-fixtures-for-backend-API-responses-3316d73d3650813ea5c8c1faa215db63)
by [Unito](https://www.unito.io)
---------
Co-authored-by: dante01yoon <bunggl@naver.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: GitHub Action <action@github.com>
## What
Adds a `cloud` Playwright project so E2E tests can run against
`DISTRIBUTION=cloud` builds, with `@cloud` / `@oss` test tagging.
## Why
100+ usages of `isCloud` / `DISTRIBUTION` across 9 categories (API
routing, UI visibility, settings, auth). Zero cloud test infrastructure
existed — cloud-specific UI components (LoginButton, SubscribeButton,
etc.) had no E2E coverage path.
## Investigation: Runtime Toggle
Investigated whether `isCloud` could be made runtime-toggleable in
dev/test mode (via `window.__FORCE_CLOUD__`). **Not feasible** —
`__DISTRIBUTION__` is a Vite `define` compile-time constant used for
dead-code elimination. Runtime override would break tree-shaking in
production.
Full investigation:
`research/architecture/cloud-runtime-toggle-investigation.md`
## What's included
### Playwright Config
- New `cloud` project alongside existing `chromium`
- Cloud project: `grep: /@cloud/` — only runs `@cloud` tagged tests
- Chromium project: `grepInvert: /@cloud/` — excludes cloud tests
### Build Script
- `npm run build:cloud` → `DISTRIBUTION=cloud vite build`
### Test Tagging Convention
```typescript
test('works in both', async () => { ... });
test('subscription button visible @cloud', async () => { ... });
test('install manager prompt @oss', async () => { ... });
```
### Example Tests
- 2 cloud-only tests validating cloud UI visibility
## NOT included (future work)
- CI workflow job for cloud tests (separate PR)
- Cloud project is opt-in — not run by default locally
## Unblocks
- Cloud-specific E2E tests for entire team
- TB-03 LoginButton, TB-04 SubscribeButton (@Kaili Yang)
- DLG-04 SignIn, DLG-06 CancelSubscription
Part of: Test Coverage Q2 Overhaul
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10546-test-infra-cloud-Playwright-project-with-cloud-oss-tagging-32f6d73d3650810ebb59dea8ce4891e9)
by [Unito](https://www.unito.io)
---------
Co-authored-by: GitHub Action <action@github.com>
Co-authored-by: Alexander Brown <drjkl@comfy.org>
## Summary
Add layout duration, style recalc duration, and heap usage metrics to CI
perf reports, while improving statistical reliability to reduce false
positive regressions.
## Changes
- **What**:
- Collect `layoutDurationMs`, `styleRecalcDurationMs`, `heapUsedBytes`
(absolute snapshot) alongside existing metrics
- Add effect size gate (`minAbsDelta`) for integer-quantized count
metrics (style recalcs, layouts, DOM nodes, event listeners) — prevents
z=7.2 false positives from e.g. 11→12 style recalcs
- Switch from mean to **median** for PR metric aggregation — robust to
outlier CI runs that dominate n=3 mean
- Increase historical baseline window from **5 to 15 runs** for more
stable σ estimates
- Reorder reported metrics: layout/style duration first (actionable),
counts and heap after (informational)
## Review Focus
The effect size gate in `classifyChange()` — it now requires both z > 2
AND absolute delta ≥ `minAbsDelta` (when configured) to flag a
regression. This addresses the core false positive issue where integer
metrics with near-zero historical variance produce extreme z-scores for
trivial changes.
Median vs mean tradeoff: median is more robust to outliers but less
sensitive to real shifts — acceptable given n=3 and CI noise levels.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10477-perf-add-layout-GC-metrics-reduce-false-positives-in-regression-detection-32d6d73d365081daa72cec96d8a07b90)
by [Unito](https://www.unito.io)
## Summary
Add historical trend visualization (ASCII sparklines + directional
arrows) to the performance PR report, showing how each metric has moved
over recent commits on main.
## Changes
- **What**: New `sparkline()`, `trendDirection()`, `trendArrow()`
functions in `perf-stats.ts`. New collapsible "Trend" section in the
perf report showing per-metric sparklines, direction indicators, and
latest values. CI workflow updated to download historical data from the
`perf-data` orphan branch and switched to `setup-frontend` action with
`pnpm exec tsx`.
## Review Focus
- The trend section only renders when ≥3 historical data points exist
(gracefully absent otherwise)
- `trendDirection()` uses a split-half mean comparison with ±10%
threshold — review whether this sensitivity is appropriate
- The `git archive` step in `pr-perf-report.yaml` is idempotent and
fails silently if no perf-history data exists yet on the perf-data
branch
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9939-feat-add-trend-visualization-with-sparklines-to-perf-report-3246d73d36508125a6fcc39612f850fe)
by [Unito](https://www.unito.io)
---------
Co-authored-by: GitHub Action <action@github.com>
Upgrades `pnpm/action-setup` from v4.2.0 to v4.4.0 across all 16
workflow files and the shared `setup-frontend` action.
## Why
GitHub Actions will force Node.js 24 as the default starting June 2,
2026. The v4.2.0 pin ran on Node.js 20 and emitted deprecation warnings
on every CI run. v4.4.0 was released specifically to address this,
updating the action runtime to Node.js 24.
- Fixes the warning: *"pnpm/action-setup@41ff72... Actions will be
forced to run with Node.js 24 by default starting June 2nd, 2026"*
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10137-ci-upgrade-pnpm-action-setup-to-v4-4-0-Node-js-24-3266d73d36508176b157fcd1d33f2274)
by [Unito](https://www.unito.io)
## Summary
Replaces two separate PR comment workflows (bundle size + performance)
with a single unified report that posts one combined comment per PR.
## Changes
- **What**: New `pr-report.yaml` aggregator workflow triggers on both
`CI: Size Data` and `CI: Performance Report` completions. Finds sibling
workflow runs by PR head SHA. Renders combined report via
`unified-report.js` (shells out to existing `size-report.js` and
`perf-report.ts`). Sections show "pending" or "failed" placeholders when
data is unavailable.
- **Breaking**: Removes `pr-size-report.yaml` and `pr-perf-report.yaml`.
Legacy `<!-- COMFYUI_FRONTEND_SIZE -->` and `<!-- COMFYUI_FRONTEND_PERF
-->` comments are auto-cleaned on first run.
- **Dependencies**: None
## Review Focus
- Concurrency key uses `head_sha` so the later-completing workflow
cancels the earlier report run, ensuring the final comment always has
both sections.
- Stale-run guard: verifies workflow_run SHA matches the live PR head
before posting.
- The `workflow_dispatch` re-trigger path from `pr-size-report.yaml` is
not carried forward — the unified workflow handles re-trigger naturally
via its dual-trigger design.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9911-feat-unified-PR-report-combining-bundle-size-and-runtime-perf-3236d73d365081baac1cce6f0d9244ac)
by [Unito](https://www.unito.io)
---------
Co-authored-by: GitHub Action <action@github.com>
## Summary
Restrict PyPI publishing of `comfyui-frontend-package` to bi-weekly
ComfyUI release cycles only, instead of every nightly version bump.
## Changes
- **What**: Move `publish_pypi` job from `release-draft-create.yaml` to
`release-biweekly-comfyui.yaml`
1. Removed `publish_pypi` job from `release-draft-create.yaml` (no
longer publishes on every merged Release PR)
2. Added `publish-pypi` job to `release-biweekly-comfyui.yaml` with tag
polling, build, publish, and PyPI availability confirmation
3. Gated `create-comfyui-pr` on `publish-pypi` success so the ComfyUI
requirements bump PR is only created after the package is confirmed
available
4. Updated ComfyUI PR body to confirm PyPI availability instead of
warning about a pending release PR
- **Breaking**: None — nightly releases still create GitHub releases and
publish npm types; only PyPI publishing timing changes
- **Dependencies**: None
## Review Focus
- The `publish-pypi` job uses `if: always() &&
needs.resolve-version.result == 'success'` to run even when
`trigger-release-if-needed` is skipped (tag already exists)
- Tag polling (30min timeout) waits for the version bump PR to be merged
before building from the tagged commit
- PyPI propagation polling (15min timeout) confirms the package is
installable before creating the ComfyUI PR
Fixes COM-16778
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9948-ops-restrict-PyPI-publishing-to-bi-weekly-ComfyUI-releases-3246d73d36508198b00fcc247ac5b58c)
by [Unito](https://www.unito.io)
---------
Co-authored-by: GitHub Action <action@github.com>
## Summary
- Adds `gh pr merge --auto --squash` after backport PR creation in the
backport workflow, so backport PRs merge automatically once checks pass
- Uses `|| echo "::warning::..."` fallback to avoid failing the workflow
if auto-merge can't be enabled (e.g. repo setting not configured)
## Test plan
- [ ] Trigger backport workflow on a test PR with `needs-backport` label
- [ ] Verify auto-merge is enabled on the created backport PR
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10108-chore-enable-auto-merge-on-backport-PRs-3256d73d3650814eb6e5fb2bdf3c5ec7)
by [Unito](https://www.unito.io)
## Summary
Remove team assignments from CODEOWNERS to reduce notification noise for
internal PRs. Add a workflow that requests team review only when
external contributors open PRs.
## Changes
- **What**: Strip `@Comfy-org/comfy_frontend_devs` and
`@Comfy-Org/comfy_maintainer` from all CODEOWNERS entries (keep
individual user assignments). Add `pr-request-team-review.yaml` workflow
that uses `pull_request_target` to request team review for
non-collaborator PRs.
- **Dependencies**: None
## Review Focus
- The workflow uses `pull_request_target` but does not check out or
execute any untrusted code — it only runs `gh pr edit --add-reviewer`.
- The `author_association` check excludes OWNER, MEMBER, and
COLLABORATOR — internal PRs will not trigger team review requests.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-10104-chore-replace-team-CODEOWNERS-with-external-PR-review-workflow-3256d73d3650813b887ac16b5e97b4c4)
by [Unito](https://www.unito.io)
## Summary
Fixes three critical issues with the CI performance reporting pipeline
that made perf reports useless on PRs (demonstrated by PR #9248 — deep
watcher removal merged without useful perf signal).
## Changes
### 1. Fix z-score baseline variance collection (`0/5 runs`)
**Root cause:** PR #9305 added z-score statistical analysis code to
`perf-report.ts`, but the historical data download step was placed in
the wrong workflow file. The report is generated in
`pr-perf-report.yaml` (a `workflow_run`-triggered job), but the
historical download was in `ci-perf-report.yaml` (the test runner) —
different runners, different filesystems.
**Fix:** Implement `perf-data` orphan branch storage:
- On push to main: save `perf-metrics.json` to `perf-data` branch with
timestamped filename
- On PR report: fetch last 5 baselines from `perf-data` branch into
`temp/perf-history/`
- Rolling window of 20 baselines, oldest pruned automatically
- Same pattern used by `github-action-benchmark` (33.7k repos)
### 2. Fix force-push comment staleness
**Root cause:** `cancel-in-progress: true` kills the perf test run
before it uploads artifacts. The downstream report workflow only
triggers on `conclusion == 'success'` — cancelled runs are ignored, so
the comment from the first successful run goes stale.
**Fix:**
- Change `cancel-in-progress: false` — with GitHub's queue depth of 1,
rapid pushes (A,B,C,D) run A and D, skipping B and C
- Add SHA validation in `pr-perf-report.yaml` — before posting, check if
the workflow_run's head SHA still matches the PR's current head. Skip
posting stale results.
### 3. Add permissions for baseline operations
- `contents: write` on CI job (needed for pushing to perf-data branch)
- `actions: read` on both workflows (needed for artifact/baseline
access)
## One-time setup required
After merging, create the `perf-data` orphan branch:
```bash
git checkout --orphan perf-data
git rm -rf .
echo '# Performance Baselines' > README.md
mkdir -p baselines
git add README.md baselines
git commit -m 'Initialize perf-data branch'
git push origin perf-data
```
The first 2 pushes to main after setup will build up variance data, and
z-scores will start appearing in PR reports (threshold is
`historical.length >= 2`).
## Testing
- YAML validated with `yaml.safe_load()`
- `perf-report.ts` `loadHistoricalReports()` already reads from
`temp/perf-history/<index>/perf-metrics.json` — no code changes needed
- All new steps use `continue-on-error: true` for graceful degradation
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9886-fix-fix-perf-CI-pipeline-z-score-baselines-force-push-staleness-baseline-storage-3226d73d365081538424c7945e71f308)
by [Unito](https://www.unito.io)
## Problem
The lint/format CI workflow was broken for fork PRs in two ways:
### 1. Node version mismatch in setup-frontend action
The `setup-frontend` shared action (created in #8377) was missed when
Node version was standardized to `.nvmrc` in #9521. It still used
`node-version: 'lts/*'` instead of `node-version-file: '.nvmrc'`.
### 2. Fork PRs with lint issues silently passed CI
Fork PRs with auto-fixable lint/format issues got a **green checkmark**
despite having unfixed issues:
1. Auto-fix steps (`lint:fix`, `format`) fix issues in the workspace
2. `Commit changes` is correctly skipped for forks (can't push to fork
branches)
3. `Final validation` passes because it runs on the already-fixed
workspace
4. The `Comment on PR about manual fix needed` step tries to post a
comment via `actions/github-script`, but fork PRs have a read-only
`GITHUB_TOKEN` — the comment silently fails (`continue-on-error: true`)
5. **Result**: workflow reports success, contributor thinks their code
is clean
## Fix
- **setup-frontend**: Use `node-version-file: '.nvmrc'` instead of
`node-version: 'lts/*'`
- **ci-lint-format**: Replace the broken fork comment step with an
explicit `exit 1` that fails CI and prints clear fix instructions in the
log. This follows the principle from `.github/AGENTS.md`: fork PRs can't
post comments, so don't try.
## Testing
- [ ] Verify fork PRs with clean code still pass
- [ ] Verify fork PRs with lint issues now properly fail (instead of
silently passing)
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9846-fix-restore-fork-PR-lint-format-CI-workflow-3226d73d3650811cb5bfe9f1f989cc0c)
by [Unito](https://www.unito.io)
## Summary
Add support for deploying full ephemeral preview environments from
frontend PRs. This is the frontend-side half — it sends `pr_number` and
`variant` (cpu/gpu) in the dispatch payload, and adds a cleanup dispatch
on PR close/unlabel.
### Changes
- **`cloud-dispatch-build.yaml`** — Add `pr_number` and `variant` to the
`frontend-asset-build` dispatch payload. Variant is derived from which
preview label triggered the event (`preview-cpu` → cpu, else gpu).
- **`cloud-dispatch-cleanup.yaml`** (new) — Fire-and-forget dispatch of
`frontend-preview-cleanup` to the cloud repo when a frontend PR is
closed or has its preview label removed. Enables synchronized teardown.
### Companion PR
Cloud-side: Comfy-Org/cloud (creates the `deploy-frontend-preview` job,
extends the reconciler)
### How it works
1. Label a frontend PR with `preview`, `preview-cpu`, or `preview-gpu`
2. Assets build and upload to GCS (existing flow)
3. Cloud deploys a full ephemeral env at `fe-pr-{N}.testenvs.comfy.org`
using all `:main` service tags
4. Subsequent pushes update the frontend SHA via AppSet upsert
5. On close/unlabel, cleanup dispatch triggers immediate teardown
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9715-feat-dispatch-frontend-PR-preview-environments-to-cloud-31f6d73d3650819da1b5ca5ce419e06e)
by [Unito](https://www.unito.io)
## Summary
Cloud build dispatch was only triggering on the `labeled` event, not on
subsequent pushes to PRs that already had a preview label.
## Changes
- **What**: Add `synchronize` to `pull_request` event types and update
the `if` condition to support all three preview labels (`preview`,
`preview-cpu`, `preview-gpu`). For `labeled` events, check the added
label name; for `synchronize` events, check existing PR labels.
## Review Focus
The `if` condition now branches on `github.event.action` to use the
correct label-checking mechanism for each event type.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9636-fix-dispatch-cloud-build-on-synchronize-for-preview-labeled-PRs-31e6d73d3650814e9069e37d6199ffc9)
by [Unito](https://www.unito.io)
## Summary
Standardize the repo's Node contract on 24 while centralizing workflow
resolution through `.nvmrc` so local setup, CI, and package metadata
stay aligned from one version file.
## Changes
- **What**: Add `package.json` `engines.node = 24.x`, switch every
`actions/setup-node` workflow in the repo to `node-version-file:
'.nvmrc'`, and update contributor and Playwright docs to point to
`.nvmrc` as the Node source of truth.
## Review Focus
The workflow behavior should be unchanged apart from sourcing the Node
version from `.nvmrc` instead of repeating literals like `20`, `22`,
`24.x`, or `lts/*`. GitHub's formatter also moved the new `engines`
block to the package metadata section near the end of `package.json`.
---------
Co-authored-by: GitHub Action <action@github.com>
Co-authored-by: Alexander Brown <drjkl@comfy.org>
## Summary
Dispatch a `frontend-asset-build` event to the cloud repo when the
`preview` label is added to a PR, so cloud can build preview assets.
## Changes
- **What**: Extended `cloud-dispatch-build.yaml` to trigger on
`pull_request` `labeled` events filtered to the `preview` label. The
payload sends the PR head SHA and branch.
## Review Focus
- The `pull_request` trigger gives a read-only `GITHUB_TOKEN`, but the
dispatch step uses `CLOUD_DISPATCH_TOKEN` so this is fine.
- Fork PRs are blocked by the existing `github.repository` guard.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9518-feat-dispatch-cloud-build-when-preview-label-is-added-to-PR-31c6d73d365081a8aab6f585960977f6)
by [Unito](https://www.unito.io)
Add PostHog as a telemetry provider for cloud builds so custom events
can be correlated with session recordings. Follows the same pattern as
MixpanelTelemetryProvider with dynamic import, event queuing, and
disabled events from remote config. Tree-shaken away in OSS builds.
The posthog-js package uses Apache-2.0 (verified from its LICENSE file)
but declares it as "SEE LICENSE IN LICENSE" in package.json, which
the license checker can't parse.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9409-feat-Add-PostHog-telemetry-provider-31a6d73d3650818b8e86c772c6551099)
by [Unito](https://www.unito.io)
## Summary
Perf report workflow fails on fork PRs because `GITHUB_TOKEN` is
read-only for forks, causing "Resource not accessible by integration" on
the PR comment step.
## Changes
- **What**: Split `ci-perf-report.yaml` into a data-collection workflow
+ a `workflow_run`-triggered reporter (`pr-perf-report.yaml`), matching
the existing `ci-size-data`/`pr-size-report` pattern. Added fork PR
permissions guidance to `.github/AGENTS.md`.
- **ci-perf-report.yaml**: Removed the `report` job and `pull-requests:
write` permission. Added PR metadata (number + base branch) artifact
upload.
- **pr-perf-report.yaml** (new): Triggered by `workflow_run` on the perf
workflow. Downloads metrics + metadata artifacts, generates report,
posts PR comment with write permissions from the default-branch context.
## Review Focus
- The two-workflow split follows the same pattern as `ci-size-data.yaml`
→ `pr-size-report.yaml`, which already works for fork PRs.
- The `workflow_run` trigger runs in the base repo context per [GitHub
Security Lab
guidance](https://securitylab.github.com/resources/github-actions-preventing-pwn-requests/),
so it safely has write permissions even for fork PRs.
- AGENTS.md guidance documents this pattern to prevent recurrence.
Fixes the failure seen in
https://github.com/Comfy-Org/ComfyUI_frontend/actions/runs/22684230751/job/65763595989?pr=9380
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9382-fix-split-perf-report-workflow-for-fork-PR-support-3196d73d365081b29b35ed354e7789e2)
by [Unito](https://www.unito.io)
## Summary
Replace fixed 10%/20% perf delta thresholds with dynamic σ-based
classification using z-scores, eliminating false alarms from naturally
noisy duration metrics (10-17% CV).
## Changes
- **What**:
- Run each perf test 3× (`--repeat-each=3`) and report the mean,
reducing single-run noise
- Download last 5 successful main branch perf artifacts to compute
historical μ/σ per metric
- Replace fixed threshold flags with z-score significance: `⚠️
regression` (z>2), `✅ neutral/improvement`, `🔇 noisy` (CV>50%)
- Add collapsible historical variance table (μ, σ, CV) to PR comment
- Graceful cold start: falls back to simple delta table until ≥2
historical runs exist
- New `scripts/perf-stats.ts` module with `computeStats`, `zScore`,
`classifyChange`
- 18 unit tests for stats functions
- **CI time impact**: ~3 min → ~5-6 min (repeat-each adds ~2 min,
historical download <10s)
## Review Focus
- The `gh api` call in the new "Download historical perf baselines"
step: it queries the last 5 successful push runs on the base branch. The
`gh` CLI is available natively on `ubuntu-latest` runners and
auto-authenticates with `GITHUB_TOKEN`.
- `getHistoricalStats` averages per-run measurements before computing
cross-run σ — this is intentional since historical artifacts may also
contain repeated measurements after this change lands.
- The `noisy` classification (CV>50%) suppresses metrics like `layouts`
that hover near 0 and have meaningless percentage swings.
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9305-feat-add-statistical-significance-to-perf-report-with-z-score-thresholds-3156d73d3650818d9360eeafd9ae7dc1)
by [Unito](https://www.unito.io)
## Summary
Adds `.github/workflows/cloud-dispatch-build.yaml` — fires a
`repository_dispatch` event (`frontend-asset-build`) to
`Comfy-Org/cloud` on push to `cloud/*` branches and `main`.
The cloud repo handles the actual build, GCS upload, and secret
management (Sentry, Algolia, GCS creds). This is fire-and-forget.
## Changes
- New workflow: `cloud-dispatch-build.yaml`
- Trigger: `push` to `cloud/*` and `main` only
- Payload: `ref` (commit SHA) + `branch` (branch name), built with `jq`
to prevent injection
- SHA-pinned `peter-evans/repository-dispatch@v4.0.1`
- Hardened: `permissions: {}`, fork guard (`if: github.repository ==
'Comfy-Org/ComfyUI_frontend'`), concurrency to avoid dispatch storms
- `cloud-deploy-frontend.yaml` left unchanged (still needed during
migration)
## Setup Required
A repository secret `CLOUD_DISPATCH_TOKEN` must be configured — see PR
description comments.
## Part of
Frontend separate deploy prep (Task 1.3)
┆Issue is synchronized with this [Notion
page](https://www.notion.so/PR-9308-feat-add-cloud-frontend-build-dispatch-workflow-3156d73d36508164a515eb968f6c5d79)
by [Unito](https://www.unito.io)
