fetch_ml/deployments/docker-compose.prod.yml

# Full Production Docker Environment with Podman and SQLite
services:
  redis:
    image: redis:7-alpine
    container_name: ml-prod-redis
    user: "999:999"
    expose:
      - "6379"
    volumes:
      - ${PROD_DATA_DIR:-./data/prod}/redis:/data
    restart: unless-stopped
    command: redis-server --appendonly yes
    healthcheck:
      test: [ "CMD", "redis-cli", "ping" ]
      interval: 30s
      timeout: 10s
      retries: 3

  api-server:
    build:
      context: .
      dockerfile: ./build/docker/secure-prod.Dockerfile
    container_name: ml-prod-api
    expose:
      - "9101"
      - "2222"
    volumes:
      - ${PROD_DATA_DIR:-./data/prod}/experiments:/app/data/experiments
      - ${PROD_DATA_DIR:-./data/prod}/active:/data/active
      - ${PROD_DATA_DIR:-./data/prod}/logs:/logs
      - ${CONFIG_DIR:-../configs}/api/multi-user.yaml:/app/configs/api/prod.yaml:ro
    depends_on:
      redis:
        condition: service_healthy
    restart: unless-stopped
    environment:
      - LOG_LEVEL=info
    healthcheck:
      test: [ "CMD", "curl", "-f", "http://localhost:9101/health" ]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s
    command: ["/bin/sh", "-c", "mkdir -p /data/active/datasets /data/active/snapshots && exec /usr/local/bin/api-server -config /app/configs/api/prod.yaml"]
    labels:
      - "traefik.enable=true"
      - "traefik.docker.network=${TRAEFIK_NETWORK:-traefik}"
      - "traefik.http.services.fetchml.loadbalancer.server.port=9101"
      - "traefik.http.routers.fetchml.rule=Host(`${FETCHML_DOMAIN}`) && (PathPrefix(`/api`) || PathPrefix(`/ws`) || Path(`/health`))"
      - "traefik.http.routers.fetchml.entrypoints=${TRAEFIK_ENTRYPOINT:-websecure}"
      - "traefik.http.routers.fetchml.tls=true"
    networks:
      - default
      - traefik

  worker:
    build:
      context: .
      dockerfile: ./build/docker/simple.Dockerfile
    container_name: ml-prod-worker
    volumes:
      - ${PROD_DATA_DIR:-./data/prod}/experiments:/app/data/experiments
      - ${PROD_DATA_DIR:-./data/prod}/active:/data/active
      - ${PROD_DATA_DIR:-./data/prod}/logs:/logs
      - ${CONFIG_DIR:-../configs}/worker/docker-prod.yaml:/app/configs/worker.yaml:ro
    depends_on:
      redis:
        condition: service_healthy
      api-server:
        condition: service_healthy
    restart: unless-stopped
    environment:
      - LOG_LEVEL=info
      - AWS_ACCESS_KEY_ID=${AWS_ACCESS_KEY_ID}
      - AWS_SECRET_ACCESS_KEY=${AWS_SECRET_ACCESS_KEY}
      - AWS_SESSION_TOKEN=${AWS_SESSION_TOKEN}
      - AWS_REGION=${AWS_REGION}
    privileged: true
    command: ["/usr/local/bin/worker", "-config", "/app/configs/worker.yaml"]

  # Smoke test profile services
  caddy:
    image: caddy:2-alpine
    container_name: ml-smoke-caddy
    profiles: ["smoke"]
    environment:
      - FETCHML_DOMAIN=localhost
    ports:
      - "8080:80"
    volumes:
      - ./Caddyfile.prod.smoke:/etc/caddy/Caddyfile:ro
      - ${SMOKE_TEST_DATA_DIR:-./data/prod-smoke}/caddy/data:/data
      - ${SMOKE_TEST_DATA_DIR:-./data/prod-smoke}/caddy/config:/config
    depends_on:
      - api-server

  ssh-test-server:
    image: linuxserver/openssh-server:latest
    container_name: ml-ssh-test
    profiles: ["smoke"]
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=America/New_York
      - PUBLIC_KEY_FILE=/tmp/test_key.pub
      - USER_NAME=test
      - PASSWORD_ACCESS=false
    volumes:
      - ./deployments/test_keys:/tmp:ro
      - ./bin/tui:/usr/local/bin/tui:ro
      - ./deployments/tui-test-config.toml:/config/.ml/config.toml:ro
    ports:
      - "2222:2222"
    depends_on:
      - caddy
      - api-server

volumes: {}

networks:
  default:
    name: ml-prod-network
  traefik:
    external: true
    name: ${TRAEFIK_NETWORK:-traefik}
  smoke-network:
    name: ml-prod-smoke-network
    profiles: ["smoke"]