services:
  backend:
    build:
      context: .
      dockerfile: Dockerfile.backend
    container_name: paddleocr-backend
    ports:
      - "8001:8000"
    environment:
      - VLLM_API_URL=http://host.docker.internal:10024/v1/chat/completions
      - VLLM_API_KEY=unused
      - VLLM_MODEL_NAME=PaddlePaddle/PaddleOCR-VL-1.5
      # RTX 6000 Blackwell Optimized (98GB VRAM)
      - MAX_GENERATION_TOKENS=4096
      - MAX_BATCH_SIZE=32
      - MAX_CONCURRENT_REQUESTS=64
      - MAX_PARALLEL_BATCHES=12
      - BATCH_JOB_PAGES_PER_BATCH=4
      - IMAGE_MAX_SIZE=2048
      - IMAGE_JPEG_QUALITY=95
      - TIMEOUT_SECONDS=60
      - MAX_NUM_SEQS=256
      - MAX_NUM_BATCHED_TOKENS=32768
    extra_hosts:
      - "host.docker.internal:host-gateway"
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8001/health"]
      interval: 30s
      timeout: 10s
      retries: 3
    networks:
      - paddle-network

  frontend:
    build:
      context: .
      dockerfile: Dockerfile.frontend
    container_name: paddleocr-frontend
    ports:
      - "8502:8501"
    environment:
      # Default backend URL (will be auto-detected from browser for LAN access)
      # Use port 8000 internally (not 8001) since containers communicate on Docker network
      - BACKEND_URL=http://backend:8000
    depends_on:
      - backend
    extra_hosts:
      - "host.docker.internal:host-gateway"
    restart: unless-stopped
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:8501 > /dev/null 2>&1"]
      interval: 30s
      timeout: 10s
      retries: 3
    networks:
      - paddle-network

networks:
  paddle-network:
    external: true
    name: paddle-network