docker-compose.yaml

version: '3.8'

services:
  # PostgreSQL Database for checkpoint/execution persistence
  postgres:
    image: postgres:16-alpine
    container_name: roma-dspy-postgres
    env_file:
      - .env
    environment:
      POSTGRES_DB: ${POSTGRES_DB:-roma_dspy}
      POSTGRES_USER: ${POSTGRES_USER:-postgres}
      POSTGRES_PASSWORD: ${POSTGRES_PASSWORD:-postgres}
    ports:
      - "${POSTGRES_PORT:-5432}:5432"
    volumes:
      - postgres_data:/var/lib/postgresql/data
      - ./docker/init-mlflow-db.sql:/docker-entrypoint-initdb.d/init-mlflow-db.sql
    healthcheck:
      test: ["CMD-SHELL", "pg_isready -U ${POSTGRES_USER:-postgres}"]
      interval: 5s
      timeout: 5s
      retries: 5
    networks:
      - roma-network
    restart: unless-stopped

  # MinIO Object Storage (S3-compatible for MLflow artifacts)
  minio:
    image: minio/minio:latest
    container_name: roma-dspy-minio
    env_file:
      - .env
    environment:
      MINIO_ROOT_USER: ${MINIO_ROOT_USER:-minioadmin}
      MINIO_ROOT_PASSWORD: ${MINIO_ROOT_PASSWORD:-minioadmin123}
    ports:
      - "${MINIO_PORT:-9000}:9000"
      - "${MINIO_CONSOLE_PORT:-9001}:9001"
    volumes:
      - minio_data:/data
    command: server /data --console-address ":9001"
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:9000/minio/health/live"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 10s
    networks:
      - roma-network
    restart: unless-stopped

  # MinIO Bucket Setup (creates mlflow bucket on startup)
  minio-setup:
    image: minio/mc:latest
    container_name: roma-dspy-minio-setup
    env_file:
      - .env
    depends_on:
      minio:
        condition: service_healthy
    entrypoint: >
      /bin/sh -c "
      mc alias set myminio http://minio:9000 ${MINIO_ROOT_USER:-minioadmin} ${MINIO_ROOT_PASSWORD:-minioadmin123};
      mc mb myminio/mlflow --ignore-existing;
      mc anonymous set download myminio/mlflow;
      exit 0;
      "
    networks:
      - roma-network

  # ROMA-DSPy API Server
  roma-api:
    build:
      context: .
      dockerfile: Dockerfile
    container_name: roma-dspy-api
    depends_on:
      postgres:
        condition: service_healthy
      minio:
        condition: service_healthy
    env_file:
      - .env
    environment:
      # Override database URL to use docker service name
      DATABASE_URL: postgresql+asyncpg://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:-postgres}@postgres:5432/${POSTGRES_DB:-roma_dspy}
      POSTGRES_ENABLED: "true"
      # MLflow tracking URI (single variable). If not set, default to in-network service.
      # To run entirely in Docker, set this to http://mlflow:5000 in your .env.
      MLFLOW_TRACKING_URI: ${MLFLOW_TRACKING_URI:-http://mlflow:5000}
      # Enable V4 traces (stores spans in database, no artifact upload needed)
      MLFLOW_ENABLE_TRACE_V4: "true"
      # S3/MinIO configuration for artifact storage
      MLFLOW_DEFAULT_ARTIFACT_ROOT: s3://mlflow
      AWS_ACCESS_KEY_ID: ${MINIO_ROOT_USER:-minioadmin}
      AWS_SECRET_ACCESS_KEY: ${MINIO_ROOT_PASSWORD:-minioadmin123}
      MLFLOW_S3_ENDPOINT_URL: http://minio:9000
      AWS_S3_ENDPOINT_URL: http://minio:9000
    ports:
      - "${API_PORT:-8000}:8000"
    volumes:
      # Application code (for development, comment out for production)
      - ./src:/app/src
      - ./config:/app/config:ro
      - ./prompt_optimization:/app/prompt_optimization
      - ./.env:/app/.env:ro  # Mount .env for optimization scripts

      # Persistent data
      - ./logs:/app/logs
      - ./executions:/app/executions
      - checkpoints_data:/app/.checkpoints
      - cache_data:/app/.cache

      # S3 mount point (mount S3 on host first, then share via volume)
      # Host must run: scripts/setup_local.sh to mount S3 to STORAGE_BASE_PATH
      - ${STORAGE_BASE_PATH:-/opt/sentient}:${STORAGE_BASE_PATH:-/opt/sentient}:rw

    networks:
      - roma-network

    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s

    restart: unless-stopped

  # MLflow Tracking Server (Optional - for experiment tracking)
  mlflow:
    build:
      context: .
      dockerfile: docker/Dockerfile.mlflow
    container_name: roma-dspy-mlflow
    env_file:
      - .env
    ports:
      - "${MLFLOW_PORT:-5000}:5000"
    environment:
      MLFLOW_BACKEND_STORE_URI: postgresql://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:-postgres}@postgres:5432/mlflow
      # Enable V4 traces (stores spans in database, no artifact upload needed)
      MLFLOW_ENABLE_TRACE_V4: "true"
      # Use S3/MinIO for artifact storage
      MLFLOW_DEFAULT_ARTIFACT_ROOT: s3://mlflow
      MLFLOW_SERVER_ALLOWED_HOSTS: "*"
      MLFLOW_SERVER_CORS_ALLOWED_ORIGINS: "*"
      # MinIO/S3 configuration
      AWS_ACCESS_KEY_ID: ${MINIO_ROOT_USER:-minioadmin}
      AWS_SECRET_ACCESS_KEY: ${MINIO_ROOT_PASSWORD:-minioadmin123}
      MLFLOW_S3_ENDPOINT_URL: http://minio:9000
      AWS_S3_ENDPOINT_URL: http://minio:9000
    networks:
      - roma-network
    command:
      - mlflow
      - server
      - --host
      - "0.0.0.0"
      - --port
      - "5000"
      - --backend-store-uri
      - postgresql://${POSTGRES_USER:-postgres}:${POSTGRES_PASSWORD:-postgres}@postgres:5432/mlflow
      - --default-artifact-root
      - s3://mlflow
      - --serve-artifacts
    depends_on:
      postgres:
        condition: service_healthy
      minio:
        condition: service_healthy
      minio-setup:
        condition: service_completed_successfully
    healthcheck:
      test: ["CMD-SHELL", "python3 -c 'import urllib.request; urllib.request.urlopen(\"http://localhost:5000/\")' || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 5
      start_period: 30s
    profiles:
      - observability  # Only start with: docker-compose --profile observability up
    restart: unless-stopped

volumes:
  postgres_data:
    driver: local
  checkpoints_data:
    driver: local
  cache_data:
    driver: local
  minio_data:
    driver: local

networks:
  roma-network:
    driver: bridge