| name | prod-up |
| description | Start the production GBot container from the published GHCR image on this host. Use when asked to bring up, start, boot, or restart the production bot. |
Bring up the production bot
Starts the single prod container with docker-compose-prod.yml: service gbot-7.0-prod, container
GBot_7.0_prod — Quart API on host port 5004 (HTTPS, self-signed) and debugpy on 5678.
⚠️ This is the live bot. It logs into Discord with the real token and reads/writes the
real Firebase RTDB, and there must only ever be one running instance. The deploy target is
the Pi (StormerPi2:/home/cgoulart/Code/GBot), whose containers are maintainer-managed —
never start it on a second host while the Pi instance is (or should be) running. To exercise
working-tree changes instead, use /qa (:qa image, watcher-safe, human-gated).
Normal deploys happen automatically: CI publishes a new arm64 image to GHCR on every code push to
develop, and the in-repo scripts/deploy-watcher.sh pulls it and redeploys (README
§Deployment). Run this skill for a deliberate manual (re)start on this host.
Steps
-
Confirm the secrets are present (compose loads/mounts both; without them the boot fails):
ls -1 Shared/gbot.env Shared/serviceAccountKey.json
If either is missing, stop and tell the user — on the Pi they live persistently in the repo dir
(gitignored); do not fabricate them.
-
Pull the published image and start (recreates the container; -d = detached):
docker compose -f docker-compose-prod.yml pull
docker compose -f docker-compose-prod.yml up -d
This runs the current GHCR image (ghcr.io/cgoulart35/gbot:latest) — the same artifact CI
publishes and the watcher auto-deploys. For a full sync to origin/develop (code + image, like
the auto-deploy), run sh scripts/deploy.sh instead. A local source build (up -d --build) is
dev/debug only — the watcher replaces it on the next published image.
-
Verify it's up:
docker compose -f docker-compose-prod.yml ps
docker compose -f docker-compose-prod.yml logs --tail=20 gbot-7.0-prod
Healthy startup: JSON log lines with GBot logged in as GBot#6890. and zero tracebacks
(#6890 is gbot.prod01. #9690 is the dev bot gbot.dev01 — seeing it here means the
wrong gbot.env is in place; see plans/completed/pi-deployment-and-updater-retirement.md.)
(common failure causes: a missing/blank required env var, or a bad FIREBASE_CONFIG_JSON).
Optional API check: curl -sk https://localhost:5004/GBot/public/leaderboard/ → 200.
Notes
- Quick restart without re-pulling (e.g. after a container hiccup):
docker compose -f docker-compose-prod.yml restart.
restart: unless-stopped is set — the container comes back on reboot/crash on its own.- Rollback:
IMAGE_TAG=<short-sha> docker compose -f docker-compose-prod.yml up -d (image tags are
the release history).
- On the local dev machine substitute
podman compose for docker compose (CLAUDE.md note).
- To stop the bot, use
/prod-down. To inspect a running bot, use /prod-logs.
