services:
  docling-serve:
    image: quay.io/docling-project/docling-serve-cu128:latest
    container_name: docling-serve
    ports:
      - "5001:5001"
    environment:
      # Enable the web UI for testing
      - DOCLING_SERVE_ENABLE_UI=true
      # CRITICAL: Required for picture description with external LLM APIs
      - DOCLING_SERVE_ENABLE_REMOTE_SERVICES=true
      # Maximum wait time for sync requests (seconds) - increase for large documents
      - DOCLING_SERVE_MAX_SYNC_WAIT=600
      # Number of local engine workers
      - DOCLING_SERVE_ENG_LOC_NUM_WORKERS=2
      # CPU thread configuration
      - OMP_NUM_THREADS=4
      - MKL_NUM_THREADS=4
      # IMPORTANT: Keep at 1 to avoid "Task Not Found" errors
      - UVICORN_WORKERS=1
    restart: unless-stopped
    # For GPU support with NVIDIA:
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [gpu]