name: rome

services:
  rome:
    image: yunfanye/rome
    build:
      context: .
      dockerfile: Dockerfile
      args:
        ROME_DOCKER_APP_CODE_MODE: ${ROME_DOCKER_APP_CODE_MODE:-source}
    ports:
      - "80:8080"
    volumes:
      - app:/app
      - user-home:/home/user
      - rome-home:/home/rome
      - ~/docker/ssh-host-keys:/etc/ssh/ssh_host_keys
      - ~/docker/tailscale:/var/lib/tailscale
    env_file: .env
    environment:
      - NODE_ENV=production
      - WEB_HOST=0.0.0.0
      - TS_HOSTNAME=rome
      - ROME_ENABLE_CHROME=1
      - ROME_CHROME_CDP_PORT=9222
      - ROME_CHROME_FULLSCREEN=0
      - ROME_CHROME_CLIPBOARD_DEFAULT_SETTING=allow
      - ROME_NOVNC_PORT=6080
      - ROME_DOCKER_APP_CODE_MODE=${ROME_DOCKER_APP_CODE_MODE:-source}
      - ROME_DOCKER_USER_MODE=${ROME_DOCKER_USER_MODE:-multi}
      # Backend → in-stack OTEL collector (otlp/http). The collector forwards
      # to the central ClickStack target using credentials from `.env`.
      - OTEL_EXPORTER_OTLP_ENDPOINT=http://otel-collector:4318
      # - SSH_USER_PASSWORD=changeme
    cap_add:
      - SYS_ADMIN
    devices:
      - /dev/fuse:/dev/fuse
    security_opt:
      - apparmor:unconfined
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "node", "-e", "fetch('http://127.0.0.1:9368/health').then(r=>{if(!r.ok)throw 1}).catch(()=>process.exit(1))"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 30s

  otel-collector:
    # contrib distribution — we need the `clickhouse` exporter, which is not
    # in the upstream `otel/opentelemetry-collector` image.
    image: otel/opentelemetry-collector-contrib:0.117.0
    container_name: rome-otel-collector
    restart: unless-stopped
    command: ["--config=/etc/otelcol/config.yaml"]
    volumes:
      - ./infra/observability/otel-collector-config.yaml:/etc/otelcol/config.yaml:ro
    # Collector-only secrets file: keeps CLICKHOUSE_* out of the rome
    # container's env (rome uses the shared `.env` above). Operators add
    # CLICKHOUSE_ENDPOINT / USERNAME / PASSWORD / DATABASE here; the
    # collector substitutes them into its config at startup via ${env:VAR}.
    env_file: .env.collector
    expose:
      - "4317" # OTLP gRPC (reserved; backend uses HTTP)
      - "4318" # OTLP HTTP
    # No `healthcheck:` here — the contrib image is distroless (no shell,
    # no curl/wget), so there's no in-container way to probe the collector's
    # /health endpoint (port 13133, exposed by the `health_check` extension
    # in otel-collector-config.yaml). Operator signal: the clickhouse
    # exporter validates its target at boot, so a broken pipeline crash-loops
    # the container and shows as `Restarting` in `docker compose ps`. Internal
    # collector logs intentionally start at WARN; tail the container for
    # configuration failures or exhausted exports. Rome does NOT depend_on the
    # collector — it must boot even if the obs stack is down (see
    # docs/architecture/observability.md).

volumes:
  app:
  user-home:
  rome-home: