A raced or interrupted previous deploy (e.g. a manual compose up overlapping the automated one) could leave a half-removed container behind, so the next recreate failed with 'removal of container ... is already in progress' (seen on the b25eb44 run, idx15). Prune stopped containers and retry once.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Sans --env-file, l'interpolation ${POSTGRES_PASSWORD:?} echoue et fait sortir
deploy.sh en non-zero (set -e) -> le job CI serait marque en echec a tort.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- docker-compose.prod.yml: postgres + redis + backend (bun) + web (nginx single-origin)
- backend/Dockerfile + entrypoint: prisma migrate deploy + seed idempotent au boot
- frontend/Dockerfile: build Vite (VITE_API_URL=https://xip.kerboul.me) servi par nginx
- deploy/nginx.conf: proxy /api + /ws vers le backend, SPA fallback
- .gitea/workflows/deploy.yml: auto-deploy SSH sur push main (runner CT121 -> CT502)
- scripts/deploy.sh: pull + rebuild de la stack
- mode open-bar (XIP_OPEN_BAR): paywall off pour tous en prod, via isFree() centralise
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
