services:
  # FastAPI 后端服务
  api:
    build:
      context: .
      dockerfile: Dockerfile
      target: api
    container_name: lnp-api
    ports:
      # 仅绑定本机，供服务器上的冒烟测试使用，不直接暴露到公网。
      - "127.0.0.1:${API_PORT:-18000}:8000"
    environment:
      - MODEL_PATH=/app/models/final/model.pt
      - RAG_POOL_CSV=/app/data/interim/internal.csv
      # 7B LLM 内部微批；外层批量筛选仍可一次提交更多条。
      - LLM_INFERENCE_BATCH_SIZE=${LLM_INFERENCE_BATCH_SIZE:-4}
    volumes:
      # 权重和 RAG 数据不进镜像，服务器上需预先放到这些路径。
      - ./models/final:/app/models/final:ro
      - ./models/mpnn:/app/models/mpnn:ro
      - ./models/qwen2.5-7b-instruct:/app/models/qwen2.5-7b-instruct:ro
      - ./data/interim/internal.csv:/app/data/interim/internal.csv:ro
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8000/"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 60s
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

  # Streamlit 前端服务
  streamlit:
    build:
      context: .
      dockerfile: Dockerfile
      target: streamlit
    container_name: lnp-streamlit
    ports:
      - "18501:8501"
    environment:
      - API_URL=http://api:8000
    depends_on:
      api:
        condition: service_started
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8501/_stcore/health"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 30s

networks:
  default:
    name: lnp-network