CI/CD Integration

Autowright fits the common automation setup: scripts fire from CI jobs or cron, while the service and fixer run somewhere long-lived. The script side is stateless — each run just needs to reach the service and artifact storage.

What the runner needs

The CI job runs your script, so it needs the script-side configuration:

Requirement	Why
serviceUrl reachable from the runner	Failure metadata is POSTed to the service
Storage endpoint reachable from the runner	The browser package uploads evidence bundles directly to S3-compatible storage
Storage credentials in env	Passed into the storage block of the config

The fixer’s secrets (ANTHROPIC_API_KEY, GITHUB_TOKEN) live with the fixer, not in CI — your runner never talks to Anthropic or opens PRs.

Honest exit codes

Autowright re-throws every error, so a failed script fails its CI job exactly as it did before. Your job status, alerting, and retry logic stay truthful — there is no “Autowright caught it, so CI went green” mode.

The fix arrives asynchronously: the job that failed stays red, and a PR (or a needs-attention notification) shows up minutes later.

Retries don’t spam PRs

Scheduled jobs retry; broken sites stay broken. Every report is deduplicated by error hash, so an hourly cron hitting the same broken selector produces one fix job — not one per run. Repeat reports within the dedupe window (default 24 hours) are counted, not re-queued.

Example: GitHub Actions on a schedule

.github/workflows/scrape.yml

name: Nightly scrape
on:
  schedule:
    - cron: '0 2 * * *'

jobs:
  scrape:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 22
      - run: npm ci
      - run: npx playwright install chromium
      - run: npx tsx src/scraper.ts
        env:
          AUTOWRIGHT_SERVICE_URL: ${{ vars.AUTOWRIGHT_SERVICE_URL }}
          STORAGE_ENDPOINT: ${{ vars.STORAGE_ENDPOINT }}
          STORAGE_BUCKET: autowright-artifacts
          MINIO_ACCESS_KEY: ${{ secrets.STORAGE_ACCESS_KEY }}
          MINIO_SECRET_KEY: ${{ secrets.STORAGE_SECRET_KEY }}

Read the env vars inside your script’s config:

autowright: {
  scriptId: 'nightly-scrape',
  serviceUrl: process.env.AUTOWRIGHT_SERVICE_URL,
  repoUrl: 'https://github.com/your-org/scrapers',
  scriptPath: 'src/scraper.ts',
  storage: {
    provider: 's3',
    endpoint: process.env.STORAGE_ENDPOINT,
    bucket: process.env.STORAGE_BUCKET,
    accessKey: process.env.MINIO_ACCESS_KEY,
    secretKey: process.env.MINIO_SECRET_KEY,
    forcePathStyle: true,
  },
}

Note

If the service is unreachable from CI, the run still captures artifacts to the runner’s local disk — but they vanish with the runner. For CI use, storage and service reachability are what make evidence durable.

Closing the loop

The PRs the fixer opens flow through your normal review process — CI runs on the fix branch like any other PR. The next scheduled run after merge is the real verification that the script healed.