Updating

A release has two moving parts: the images (ghcr.io/batazor/autopilot/bot:latest, web:latest) and the docker-compose.prod.yml itself. Pulling only the images is the most common mistake — the compose file evolves too (health checks, services, env), and your copy is a static download that nothing refreshes for you.

The correct update

1. Refresh the compose file. It changes between releases, and your local copy never updates itself:

   curl -fsSL https://batazor.github.io/autopilot-page/docker-compose.prod.yml -o docker-compose.prod.yml

2. Pull the new images and recreate the containers:

   docker compose -f docker-compose.prod.yml up -d --pull always

3. Check everything is healthy:

   docker compose -f docker-compose.prod.yml ps

That’s the happy path. If a container comes back unhealthy or looks stuck on the old version, read the next section — it’s almost always this.

Trap: --pull always won’t fix a stuck container

A re-run over a broken container fast-fails

up --pull always only recreates a container when the image digest actually changed. If an earlier run left a container unhealthy, Compose keeps it and up fast-fails almost instantly with “dependency … is unhealthy” — it does not rebuild the broken one. The classic symptom:

✘ Container autopilot-api   Error   dependency failed to start
dependency failed to start: container autopilot-api is unhealthy

Force a clean recreation with an explicit down first:

docker compose -f docker-compose.prod.yml down
docker compose -f docker-compose.prod.yml up -d --pull always

down removes the containers and the network but keeps all your data — your accounts are untouched.

Start fresh

Throw everything away — accounts, state, cache — and bring the stack back up empty:

docker compose -f docker-compose.prod.yml down -v
docker compose -f docker-compose.prod.yml up -d --pull always

down -v is irreversible

It deletes all your data for good. Afterwards you’ll re-add your accounts — so use it only when you really want a clean start.

When something is still off

1. Read the real error. A failing health probe tells you nothing; the logs do:

   docker logs autopilot-api
   docker logs autopilot-bot

2. macOS / Windows — re-check Host networking. If the API never reaches healthy after an update, make sure Host networking is still enabled (Docker Desktop → Settings → Resources → Network). The stack relies on it — see Images & networking.

Still stuck? Head to Troubleshooting.