services:
  aftershock:
    build: .
    image: aftershock:latest
    expose:
      - "8788"
    ports:
      # Default loopback-only (prod sits behind Caddy). A LAN/staging box can set
      # OBSERVATORY_BIND=0.0.0.0 in its .env to reach the app directly at host:8788.
      - "${OBSERVATORY_BIND:-127.0.0.1}:8788:8788"
    env_file:
      # Optional: copy .env.example to .env for live LLM runs (DASHSCOPE_API_KEY) and the
      # observatory token. A fresh clone runs without it (scripted arm + recorded-run replay).
      - path: .env
        required: false
    environment:
      OBSERVATORY_ORIGIN: ${OBSERVATORY_ORIGIN:-http://127.0.0.1:8788}
    volumes:
      - aftershock-runs:/data/runs
    restart: unless-stopped
    # Liveness probe: /api/status is cheap (no file IO) and hangs whenever the single
    # event loop is blocked — the failure mode behind the 2026-06-25 "No runs found"
    # outage (a 62.5s /api/runs scan starved every endpoint). Marks the container
    # unhealthy; the `autoheal` sidecar then restarts it (Compose's restart policy acts
    # only on EXIT, not on a hung-but-running app). Uses the app's own Python — the
    # slim image has no curl/wget. start_period covers boot + demo-run seeding.
    healthcheck:
      test:
        - CMD
        - uv
        - run
        - --no-sync
        - python
        - -c
        - import urllib.request,sys; sys.exit(0 if urllib.request.urlopen('http://127.0.0.1:8788/api/status', timeout=5).getcode()==200 else 1)
      interval: 30s
      timeout: 8s
      retries: 3
      start_period: 45s
    labels:
      autoheal: "true"

  # Restarts any container labelled autoheal=true the moment its healthcheck goes
  # unhealthy — Docker Compose's restart policy only reacts to a container EXITING, not
  # to a process that is still running but wedged. Needs read/write on the Docker socket
  # to issue the restart, so it is effectively root on the host: acceptable on this
  # single-tenant demo box. Pin `latest` → a digest for stricter supply-chain hygiene.
  autoheal:
    image: willfarrell/autoheal:latest
    restart: unless-stopped
    environment:
      AUTOHEAL_CONTAINER_LABEL: autoheal
      AUTOHEAL_INTERVAL: "10"
      AUTOHEAL_START_PERIOD: "60"
      AUTOHEAL_DEFAULT_STOP_TIMEOUT: "20"
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock

  caddy:
    image: caddy:2-alpine
    depends_on:
      - aftershock
    ports:
      - "80:80"
      - "443:443"
    environment:
      AFTERSHOCK_HOSTNAME: ${AFTERSHOCK_HOSTNAME:-localhost}
    volumes:
      - ./deploy/Caddyfile:/etc/caddy/Caddyfile:ro,Z
      - caddy-data:/data
      - caddy-config:/config
    restart: unless-stopped

volumes:
  aftershock-runs:
  caddy-data:
  caddy-config: