CI
The daily refresh and the deploy, on GitHub Actions.
- Refresh (
.github/workflows/refresh.yml) runs daily at 03:00 UTC: check, crawl, and on the 1st of the month discover new shops. It starts from the previous crawl data, so a shop that fails keeps its records, publishesgenerated/as thegenerated.tar.gzasset of thedatarelease, then runs Deploy. - Deploy (
.github/workflows/deploy.yml) runs on every push tomain: downloads the latest crawl data, builds, runs the end-to-end tests and publishes to GitHub Pages. Without crawl data, or when a test fails, it stops, so the live site keeps the previous build (the Playwright report is attached to the failed run). - Health report (
npm run health,scripts/health/, a step of Deploy): compares the build with the last one of an earlier day (records, searchable shops, site size, crawl failures and age, end-to-end tests), thresholds indata/config.jsonhealth. Past numbers are kept ashealth.jsonin thedatarelease. When degraded it opens or updates one issue labelledhealth; once healthy again it comments and closes it. - Run either by hand from the Actions tab (Refresh has a “discover” option).