Sign inSign up

simonjoz/runpod-comfy-worker

By simonjoz

Updated about 1 year ago

RunPod worker image with pre-installed ComfyUI, custom nodes and automations.

Image
Machine learning & AI
0

504

simonjoz/runpod-comfy-worker repository overview

RunPod Comfy Worker

Docker Pulls Docker Stars Docker Image Size Docker Image Version

This Docker image is built on top of the runpod-comfy-base image and provides a ready-to-use ComfyUI environment optimized for RunPod..

Features

  • Preinstalled ComfyUI (depends on image - base: 0.3.32)
  • CUDA support (depends on image - base: 12.6.3)
  • Memory optimization via tcmalloc
  • Auto-starts ComfyUI, Jupyter Lab, and Filebrowser
  • Installs a curated set of ComfyUI custom nodes
  • Optional model downloads based on environment config
  • Optional SSH access via public key injection

Exposed Ports

Container PortServiceNotes
3000ComfyUIUI at http://<host>:3000
8888Jupyter LabToken = JUPYTER_PASSWORD
4040FilebrowserAdmin = FB_USERNAME / FB_PASSWORD
22SSH (optional)Enabled if PUBLIC_KEY is provided

Environment Variables

VariableDescriptionDefault
NETWORK_VOLUMEPath to volume disk storage for models (e.g., /workspace)unset
PUBLIC_KEYInjects SSH key for remote accessunset
JUPYTER_PASSWORDRequired for Jupyter Lab access (used as login token)unset
FB_USERNAMERequired Filebrowser admin usernameunset
FB_PASSWORDRequired Filebrowser admin passwordunset
CUSTOM_NODES_LISTComma separated list of Custom Nodes to install (e.g., [email protected],rgthree-comfy)unset
DOWNLOAD_SDXLIf true, download SDXL modelsfalse
DOWNLOAD_FLUXIf true, download Flux.1 Dev modelsfalse
DOWNLOAD_FLUX_KONTEXTIf true, download Flux.1 Dev Kontext modelsfalse
DOWNLOAD_WANIf true, download Wan 2.1 modelsfalse
DOWNLOAD_HUNYUAN_DITIf true, download Hunyuan DiT modelsfalse

Default Custom Nodes List

Note: The CUSTOM_NODES_LIST environment variable overwrites the default list. It is useful if only a subset of nodes is needed, or if you want to define your own custom nodes. Please note that the default list can take some time to install.

Usage

Run the container with the required ports exposed. Volume mounts and environment variables are optional, depending on your use case.

Basic
docker pull simonjoz/runpod-comfy-worker:<tag>
docker run --gpus all -p 3000:3000 -p 8888:8888 -p 4040:4040 simonjoz/runpod-comfy-worker:<tag>
Advanced
docker run --gpus all \
  -v /host/dir:/workspace \
  -e NETWORK_VOLUME=/workspace \
  -e PUBLIC_KEY="$(cat ~/.ssh/id_rsa.pub)" \
  -e JUPYTER_PASSWORD=<pw> \
  -e FB_USERNAME=<user> \
  -e FB_PASSWORD=<pw> \
  -e DOWNLOAD_SDXL=true \
  -e DOWNLOAD_FLUX=true \
  -e DOWNLOAD_FLUX_KONTEXT=true \
  -e DOWNLOAD_WAN=true \
  -e DOWNLOAD_HUNYUAN_DIT=true \
  -p 3000:3000 \
  -p 8888:8888 \
  -p 4040:4040 \
  -p 22:22 \
  simonjoz/runpod-comfy-worker:<tag>
Docker Compose
version: '3.8'

services:
  comfy-worker:
    image: simonjoz/runpod-comfy-worker:<tag>
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: all
              capabilities: [ gpu ]
    volumes:
      - /host/dir:/workspace
    environment:
      NETWORK_VOLUME: /workspace
      PUBLIC_KEY: "${PUBLIC_KEY}"
      JUPYTER_PASSWORD: "${JUPYTER_PASSWORD}"
      FB_USERNAME: "${FB_USERNAME}"
      FB_PASSWORD: "${FB_PASSWORD}"
      DOWNLOAD_SDXL: "true"
      DOWNLOAD_FLUX: "true"
      DOWNLOAD_FLUX_KONTEXT: "true"
      DOWNLOAD_WAN: "true"
      DOWNLOAD_HUNYUAN_DIT: "true"
    ports:
      - "3000:3000"
      - "8888:8888"
      - "4040:4040"
      - "22:22"

License

MIT

Tag summary

Content type

Image

Digest

sha256:eed2db02b

Size

11.2 GB

Last updated

about 1 year ago

docker pull simonjoz/runpod-comfy-worker:0.1.1-cuda12.6.3-ubuntu24.04