From a failing test to a shipped fix, automatically

The failure → finding → fix → retest loop

Every Playwright run lands in OMG IQ and travels four labeled stages in this exact order. Each one is a different state of the same Finding, visible to humans and agents on the same board:

Integration steps — setup & usage

Wiring Playwright into the QA / hardening workflow takes two passes: a one-time setup (create a project + API key, install the reporter, point it at OMG IQ), then the usage loop on every CI run (the reporter ships the run to the ingestion API and OMG IQ does the rest).

Setup — one time

1. Create the project (or pick one). In OMG IQ, every Finding belongs to a project; the reporter targets one project per run. If you don't have a project yet, sign in and create one — the project id (UUID) and key (e.g. CHK) are what the reporter needs.
2. Generate a project-scoped API key. Under Account → API keys, create a new key, bound to that project, with the scopes your agent needs (read + write, or read-only for a watcher). Store it as a CI secret — e.g. OMGQA_API_KEY in GitHub Actions, GitLab CI, or Azure DevOps.
3. Install the reporter. Add @omgqa/playwright-reporter to your test project's devDependencies:
   # npm npm install --save-dev @omgqa/playwright-reporter # or pnpm / yarn / bun pnpm add -D @omgqa/playwright-reporter
   For other runners: @omgqa/cypress-reporter for Cypress, @omgqa/jest-reporter for Jest/Vitest, @omgqa/pytest-reporter for pytest. The shape of every reporter's payload is identical.
4. Point the reporter at OMG IQ. Add it as a Playwright reporter. The reporter fails soft — a misconfigured or unreachable OMG IQ will never fail your test run — so onboarding is safe to leave on from the first commit:
   // playwright.config.ts export default defineConfig({ reporter: [ ['list'], ['@omgqa/playwright-reporter', { baseUrl: 'https://api.omgiq.com', apiKey: process.env.OMGQA_API_KEY, // create under Account → API keys projectId: '<your-project-id>', }], ], });
   No other config is required to get the failure → finding → fix → retest loop live. All other options (branch, commit, dedupe, timeoutMs, softFail) have sensible defaults.
5. (Optional) wire ingest secrets in CI. Set OMGQA_API_KEY and OMGQA_PROJECT_ID as masked CI variables. If you want the QA / hardening workflow to also pick up coverage and JUnit XML, add OMGQA_COVERAGE with the path to a Cobertura or lcov report and OMGQA_JUNIT with the JUnit XML path.

Usage — every CI run

1. Tests run as normal. npx playwright test / pnpm test / CI trigger — nothing changes about how Playwright itself runs.
2. Reporter publishes the run. After the suite, the reporter POSTs the run JSON to /api/v1/ingest/playwright-mcp?projectId=…. Failures are turned into Findings, deduped by signature, and decorated with trace/video/screenshot when available.
3. An agent (or you) picks up the Finding. Triage happens through the same MCP tools the rest of the system uses — search_findings, get_finding_details, find_similar_findings — so an agent on the team can drive the same board you see in the UI.
4. Open a PR linked to the Finding. Either by hand (paste the CHK-… key in the PR body — OMG IQ links it on merge) or through the agent via link_pr. The PR's commit and merge event keep the Finding's history intact.
5. On merge, the Finding moves to Ready for retest. OMG IQ moves the state when the linked PR merges; you don't have to remember. The retest Playwright run on the merge commit is what a human uses to close it.
6. A human disposition closes the loop. Fixed → Resolved. Won't fix / accepted risk / duplicate → closed with the reason captured for the record. Nothing in the fleet crosses that gate on its own.

What attaches to a Finding

Every Playwright run ships with rich evidence that lands automatically:

• Trace — the step-by-step Playwright trace, viewable in-app.
• Video — the full run recording.
• Screenshot — captured at the moment of failure.
• HAR — network log when the failure is HTTP-shaped.
• Logs — the console / page errors that triggered the failure.

Files upload directly to private storage via short-lived signed URLs — the API never proxies bytes. In the app you can open a trace or HAR right inside the Finding on one synchronized timeline.

Flakiness & root-cause signals, included

Because every Playwright run lands in OMG IQ, the platform knows what your tests actually do — not just what the latest run says.

What a fix looks like end-to-end

1. Failure. Playwright runs npm test on the PR. Two tests fail on Safari; the reporter ships the run to OMG IQ.
2. Finding. The ingestion layer turns each failure into a Finding (or bumps an existing one if it's a repeat) with the trace and screenshots attached.
3. Fix. An agent claims the checkout Finding, opens a Story (CHX-87), branches feat/CHX-fix-checkout, writes the fix, runs make verify green in its isolated worktree, pushes the branch, opens a PR, and hands it to the merge lane. CHX-87 reaches Ready for Acceptance.
4. Retest. The PR merges, the Finding moves to Ready for retest, and the same Playwright suite runs against the merge commit. Both tests pass. A reviewer (you) compares the new trace against the original failure, confirms the fix, and closes the Finding.