Skip to main content

ComfyUV Deployment Guide

Overview

This guide explains how to deploy the Docker image:

sandichhuu/comfyuv on Docker Hub

TheAn optimized, production-ready ComfyUI Docker image providesbuilt awith uv for ultra-fast Python package management. It comes pre-configured environmentwith modern performance optimizations like Triton, SageAttention.

Important Note:

⚠️ This image is only support nvidia RTX30xx (sm86).

⚠️ Unsupport xformers (cu132 have no pre-build for runningthis ComfyUIpackage).

with:

Why use this Docker image for ComfyUI?

    Blazing Fast Execution: Leverages uv for ultra-fast package management and bytecode compilation. Built on CUDA 13.2, delivering superior performance compared to older versions like CUDA 12.8 or 13.0. Precompiled Ecosystem: Comes out-of-the-box with precompiled essential packages and popular custom nodes, saving you significant setup and compile time. Privacy-Focused (No Analytics): Completely stripped of telemetry and analytics tracking. Your data and workflows remain entirely private. Optimized for cloud: You can public the website without configuration. Use caddy or nginx for authentication.

    Preinstalled sage-attn and custom nodes:

    deploymentComfy-Org/Nvidia_RTX_Nodes_ComfyUI workflowComfyUI-KJNodes

    RequirementsQuick Start (Docker Compose)

    Required

    The

      easiest NVIDIAway GPUto run comfyuv is using Docker Compose with NVIDIA Container Toolkit Dockerinstalled. Engine

      ⚠️ CUDA-compatibleMake driverssure

      nvidia driver,
        cuda 16GB+tookit VRAMare SSDinstalled storagebefore Linuxstart host(just run
        nvidia-smi

        Quickon Start

        terminal

        Pullto thetest). Image

        Create 

        a dockerdocker-compose.yml pull sandichhuu/comfyuv:latest

        Docker Compose Example

        file: 
        services:
  comfyuv:
    image: sandichhuu/comfyuv:cu132cpp313sm86cu132
    container_name: comfyuv
    ipc: host
    port:
      8188:8188
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]
    volumes:
      - ./config.ini:/comfy/user/__manager/config.ini
      - ./input:/comfy/input
      - ./output:/comfy/output
      - ./workflows:/comfy/user/default/workflows
      - ./models:/comfy/models
      - comfy:./workflows:/comfy/user/default/workflows
      
      - comfyuv:/comfy
      - comfyuv_python_packages:uv_cache:/usr/local/lib/python3.13/site-packages
    environment:
      - DO_NOT_TRACK=1
      - COMFY_NO_TELEMETRY=1root/.cache/uv
    command: >
      --listen 0.0.0.0
      --enable-triton-backend
#     --enable-manageruse-flash-attention unsupport becase it take long times to compile.
      --use-sage-attention
      --disable-pinned-memory
      --fast fp8_matrix_mult autotune
      --front-end-version Comfy-Org/ComfyUI_frontend@latest

volumes:
  comfyuv:
  comfyuv_python_packages:uv_cache:

        Start:

        Structure

        dockercomfyuv/
compose├── updocker-compose.yml
-d├──
        config.ini
        ├──

        Access ComfyUI

        Open:

        http://localhost:8188

        project/input/
├── models/
├── input/workflows/
├└── output/
├── workflows/
├── config.ini
└── docker-compose.yml

        For long-term deployments:

          Reverse proxy with Caddy or Traefik Modify config.ini 
          [default]
git_exe = 
use_uv = True
use_unified_resolver = False
channel_url = https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main
share_option = all
bypass_ssl = False
file_logging = True
update_policy = stable-comfyui
windows_selector_event_loop_policy = False
model_download_by_agent = False
downgrade_blacklist = 
security_level = normal
always_lazy_install = False
network_mode = personal_cloud
db_mode = cache
verbose = False

          Notes

            First startup may take time due to model indexing. Large models should be mounted from persistent storage. SageAttention support depends on GPU architecture and CUDA compatibility. Some custom nodes may still require additional dependencies.

            (docs.docker.com)

            Back to top