pfm-ocr/backend/docker-compose.yml

name: ai-ocr-pfm-2026

services:
  nginx:
    image: nginx:alpine
    container_name: paddleocr-nginx
    ports:
      - "${APP_PORT:-8000}:80"
    volumes:
      - ./nginx.conf:/etc/nginx/nginx.conf:ro
    depends_on:
      - vllm-server
      - pipeline-api
      - gradio-ui
      - pfm-web-app
    restart: unless-stopped

  vllm-server:
    build:
      context: .
      target: vllm-server
    container_name: paddleocr-vllm-server
    image: paddleocr-vllm-server:latest
    environment:
      - GENAI_HOST=0.0.0.0
      - GENAI_PORT=8118
      - GENAI_MODEL=${GENAI_MODEL:-PaddleOCR-VL-1.6-0.9B}
      - GENAI_BACKEND=${GENAI_BACKEND:-vllm}
      - VLLM_CONFIG=${VLLM_CONFIG:-config/vllm_config.yaml}
      - CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES:-0}
      - PADDLE_PDX_DISABLE_MODEL_SOURCE_CHECK=True
    # No exposed ports; internal only
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    volumes:
      - hf_cache:/root/.cache/huggingface
      - paddle_cache:/root/.paddleocr
      - paddlex_cache:/root/.paddlex
      - ./config:/app/config
      - ./.env:/app/.env:ro
    restart: unless-stopped

  pipeline-api:
    build:
      context: .
      target: pipeline-api
    container_name: paddleocr-pipeline-api-v10
    image: paddleocr-pipeline-api:latest
    environment:
      - PIPELINE_CONFIG=${PIPELINE_CONFIG:-config/pipeline_config_vllm.yaml}
      - PIPELINE_HOST=0.0.0.0
      - PIPELINE_PORT=8090
      - PIPELINE_DEVICE=gpu:0
      - CUDA_VISIBLE_DEVICES=${CUDA_VISIBLE_DEVICES:-0}
      - VLLM_SERVER_URL=http://paddleocr-pfm-web-app:3000/api/vllm-proxy/v1
      - PADDLE_PDX_DISABLE_MODEL_SOURCE_CHECK=True
    # No exposed ports; internal only
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]
    volumes:
      - paddle_cache:/root/.paddleocr
      - paddlex_cache:/root/.paddlex
      - ./config:/app/config
      - ./pfm-web-app/public/produk-pfm/models:/app/pfm-web-app/public/produk-pfm/models:ro
      - ./.env:/app/.env:ro
    depends_on:
      - vllm-server
    restart: unless-stopped

  gradio-ui:
    build:
      context: .
      target: gradio-ui
    container_name: paddleocr-gradio-ui
    image: paddleocr-gradio-ui:latest
    environment:
      - API_URL=http://paddleocr-pipeline-api-v10:8090/layout-parsing
      - GRADIO_PORT=7870
      - GRADIO_MCP_SERVER=True
    # No exposed ports; internal only
    depends_on:
      - pipeline-api
    restart: unless-stopped

  pfm-web-app:
    build:
      context: .
      target: pfm-web-app
    container_name: paddleocr-pfm-web-app
    image: paddleocr-pfm-web-app:latest
    command: npm run dev
    environment:
      - PIPELINE_URL=http://paddleocr-pipeline-api-v10:8090/layout-parsing
      - NODE_ENV=development
      - PGHOST=paddleocr-db
      - PGPORT=5432
      - PGUSER=postgres
      - PGPASSWORD=postgres
      - PGDATABASE=dopfm
      - JWT_SECRET=${JWT_SECRET:-dev-only-insecure-secret-change-me}
    pid: "host"
    volumes:
      - ./pfm-web-app:/app
      - /app/node_modules
      - /app/.next
      - ./uploads:/uploads
      - /var/run/docker.sock:/var/run/docker.sock
    # No exposed ports; internal only
    depends_on:
      - pipeline-api
      - db
    extra_hosts:
      - "host.docker.internal:host-gateway"
    restart: unless-stopped

  db:
    image: postgres:15-alpine
    container_name: paddleocr-db
    environment:
      - POSTGRES_USER=postgres
      - POSTGRES_PASSWORD=postgres
      - POSTGRES_DB=dopfm
    volumes:
      - pgdata:/var/lib/postgresql/data
      - ./db/migrations:/docker-entrypoint-initdb.d:ro
    restart: unless-stopped

volumes:
  hf_cache:
    name: paddleocr_hf_cache
  paddle_cache:
    name: paddleocr_paddle_cache
  paddlex_cache:
    name: paddleocr_paddlex_cache
  pgdata:
    name: paddleocr_pgdata