docker-compose.yml

services:
  mlflow:
    # MLflow tracking server:
    # - stores experiment runs
    # - stores logged artifacts/model metadata
    build:
      context: .
      dockerfile: docker/Dockerfile.mlflow
    ports:
      # Host:Container
      - "5000:5000"
    volumes:
      # Persist MLflow database/artifacts across restarts.
      - mlflow:/mlflow
    # Automatically restart unless explicitly stopped by an operator.
    restart: unless-stopped
    healthcheck:
      # Healthcheck lets Docker know if the service is actually responding.
      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:5000/').read()"]
      interval: 10s
      timeout: 3s
      retries: 10

  api:
    # FastAPI inference service.
    build:
      context: .
      dockerfile: docker/Dockerfile.api
    environment:
      # API uses MLflow for model loading/registration and local data volume for logs.
      MLFLOW_TRACKING_URI: http://mlflow:5000
      MLFLOW_MODEL_URI: ${MLFLOW_MODEL_URI:-}
      MODEL_PATH: ${MODEL_PATH:-}
      PREDICTION_LOG_PATH: /app/data/predictions.jsonl
      API_KEY: ${API_KEY:-}
      LOG_LEVEL: ${LOG_LEVEL:-INFO}
      MAX_UPLOAD_MB: ${MAX_UPLOAD_MB:-10}
    ports:
      - "8000:8000"
    volumes:
      # Mount local data folder so predictions/reference files are accessible on the host.
      - ./data:/app/data
    depends_on:
      - mlflow
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "python", "-c", "import urllib.request; urllib.request.urlopen('http://localhost:8000/health').read()"]
      interval: 10s
      timeout: 3s
      retries: 10

  nginx:
    # Public entrypoint / reverse proxy in front of the API.
    image: nginx:1.27-alpine
    ports:
      - "8080:80"
    volumes:
      # Inject our custom proxy config.
      - ./nginx/nginx.conf:/etc/nginx/nginx.conf:ro
    depends_on:
      - api
    restart: unless-stopped

  prometheus:
    # Metrics collector. Periodically scrapes /metrics from the API.
    image: prom/prometheus:v2.55.1
    ports:
      - "9090:9090"
    volumes:
      - ./monitoring/prometheus.yml:/etc/prometheus/prometheus.yml:ro
      - prometheus:/prometheus
    depends_on:
      - api
    restart: unless-stopped

  grafana:
    # Dashboard UI on top of Prometheus metrics.
    image: grafana/grafana:11.5.2
    ports:
      - "3000:3000"
    environment:
      # Default local login for development.
      GF_SECURITY_ADMIN_USER: admin
      GF_SECURITY_ADMIN_PASSWORD: admin
    volumes:
      # Persist dashboards/settings and auto-provision datasource config.
      - grafana:/var/lib/grafana
      - ./monitoring/grafana/provisioning:/etc/grafana/provisioning:ro
    depends_on:
      - prometheus
    restart: unless-stopped

volumes:
  # Named volumes survive container recreation.
  mlflow: {}
  prometheus: {}
  grafana: {}