# NVIDIA CUDA deployment

services:
  docling-serve:
    image: ghcr.io/docling-project/docling-serve-cu126:main
    container_name: docling-serve
    ports:
      - "5001:5001"
    environment:
      DOCLING_SERVE_ENABLE_UI: "true"
      NVIDIA_VISIBLE_DEVICES: "all" # https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/docker-specialized.html
    # deploy:  # This section is for compatibility with Swarm
    #   resources:
    #     reservations:
    #       devices:
    #         - driver: nvidia
    #           count: all
    #           capabilities: [gpu]
    runtime: nvidia
    restart: always