| name | qa-smoke |
| description | Run Crucible's full QA smoke suite before merging — backend pytest, frontend typecheck+lint, and the Playwright E2E journeys — and report pass/fail per stage. Use when asked to smoke-test, QA, or verify the app end to end before a merge/release. |
QA smoke run
Run these three stages in order, from the repo root. Report pass/fail per stage;
do not stop at the first failure unless a stage is a hard prerequisite for the
next (noted below).
1. Backend tests
source venv/bin/activate && python -m pytest backend/tests/ -q
Same-shell venv rule: never wrap the source in ( … ) — activation is
discarded and it runs on system Python (no fastapi/sqlalchemy).
2. Frontend typecheck + lint (hard prerequisite for stage 3)
cd frontend && npm run build && npm run typecheck:e2e && npm run lint
npm run build (tsc -b && vite build) is the only real typecheck for src/
AND it refreshes frontend/dist, which the E2E server serves. Stale dist tests
stale code, so a green build here is required before stage 3. e2e/ sits
outside the solution config, so tsc -b never sees it — npm run typecheck:e2e
is the check for the Playwright specs (~2 s). Lint currently carries
pre-existing react-hooks debt (non-blocking in CI) — report new lint errors your
change introduced, but pre-existing hits are not a stage failure.
3. Playwright E2E
cd frontend
npx playwright install chromium
npx playwright test
The config starts its own backend on :8199 against a throwaway DB
(e2e/serve.sh) — it never touches the live repo-root DB. On failure, the HTML
report is written to frontend/playwright-report/; open it to triage.
Report
Summarise each stage: backend pytest: N passed, frontend build+lint: ok,
e2e: N passed / M failed. On any failure, quote the failing test name and the
assertion, and point at playwright-report/ for E2E.
Pitfalls
- Port 8199 already bound (a previous run's server, or a dev session on it) →
the webServer step fails to start. Kill the stray process first.
- Never point the suite at a real dev server.
reuseExistingServer is false
precisely so it always spins up its own throwaway backend.
- Don't rebuild the frontend mid-run. Rebuild only between full runs, in
stage 2 — never while
playwright test is executing against the served dist.
- E2E deliberately excludes Restart/Shutdown, ComfyUI, booru and all GPU jobs
(no ML server during e2e); do not add journeys that need them.
