services:
  paddleocr-vl-api:
    build:
      context: .
      dockerfile: gateway.Dockerfile
      args:
        HPS_SDK_DIR: ${HPS_SDK_DIR:-paddlex_hps_PaddleOCR-VL-1.6_sdk}
    container_name: paddleocr-vl-api
    ports:
      - 8080:8080
    depends_on:
      paddleocr-vl-pipeline:
        condition: service_healthy
    environment:
      - HPS_TRITON_URL=paddleocr-vl-pipeline:8001
      - HPS_VLM_URL=${HPS_VLM_URL:-http://paddleocr-vlm-server:8080}
      - HPS_MAX_CONCURRENT_INFERENCE_REQUESTS=${HPS_MAX_CONCURRENT_INFERENCE_REQUESTS:-16}
      - HPS_MAX_CONCURRENT_NON_INFERENCE_REQUESTS=${HPS_MAX_CONCURRENT_NON_INFERENCE_REQUESTS:-64}
      - HPS_INFERENCE_TIMEOUT=${HPS_INFERENCE_TIMEOUT:-600}
      - HPS_LOG_LEVEL=${HPS_LOG_LEVEL:-INFO}
      - HPS_UVICORN_WORKERS=${HPS_UVICORN_WORKERS:-4}
    restart: unless-stopped
    healthcheck:
      test: ["CMD-SHELL", "curl -f http://localhost:8080/health || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 3

  paddleocr-vl-pipeline:
    build:
      context: .
      dockerfile: pipeline.Dockerfile
      args:
        BASE_IMAGE: ${HPS_TRITON_BASE_IMAGE:-ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlex/hps:paddlex${HPS_PADDLEX_VERSION:-3.6}-gpu}
        DEVICE_TYPE: ${HPS_DEVICE_TYPE:-gpu}
        HPS_SDK_DIR: ${HPS_SDK_DIR:-paddlex_hps_PaddleOCR-VL-1.6_sdk}
    container_name: paddleocr-vl-pipeline
    depends_on:
      paddleocr-vlm-server:
        condition: service_healthy
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ["${HPS_DEVICE_ID:-0}"]
              capabilities: [gpu]
    shm_size: 4gb
    restart: unless-stopped
    healthcheck:
      test: ["CMD-SHELL", "curl -f http://localhost:8000/v2/health/ready || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 3
      start_period: 60s

  paddleocr-vlm-server:
    image: ccr-2vdh3abv-pub.cnc.bj.baidubce.com/paddlepaddle/paddleocr-genai-vllm-server:latest-nvidia-gpu
    container_name: paddleocr-vlm-server
    volumes:
      - ./${HPS_SDK_DIR:-paddlex_hps_PaddleOCR-VL-1.6_sdk}/server/pipeline_config.yaml:/config/pipeline_config.yaml:ro
      - ./genai_server_entrypoint.sh:/entrypoint.sh:ro
    command: ["/bin/sh", "/entrypoint.sh"]
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              device_ids: ["${HPS_DEVICE_ID:-0}"]
              capabilities: [gpu]
    restart: unless-stopped
    healthcheck:
      test: ["CMD-SHELL", "curl -f http://localhost:8080/health || exit 1"]
      interval: 10s
      timeout: 5s
      retries: 3
      start_period: 300s