ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
ComfyUI on a separate Windows GPU host
Run this path when a different device provides LiteLLM or other ODS services.
It installs only ComfyUI in the ods-comfyui-standalone Docker Compose project.
It does not install Portal, Pixel, a language model, or a dashboard. A full ODS
installation on the same Windows computer is left alone.
From the ODS source directory, in a normal PowerShell window:
.\ods\installers\windows\install-comfyui-standalone.ps1 -DryRun
.\ods\installers\windows\install-comfyui-standalone.ps1
Use -DataRoot C:\path\with\space\ComfyUI to keep models and output on a
different local drive, or -Port 8190 if 8188 is occupied. The installer checks
for an NVIDIA GPU, a local Docker Desktop Linux engine, GPU access from a CUDA
container, free disk, and a free loopback port. A CUDA/PyTorch build needs at
least 25 GiB free on the Docker Desktop and data drives before it begins.
After a healthy install, open http://127.0.0.1:8188/ (or the chosen port).
The initial install contains no image checkpoint. Put a compatible model in
<DataRoot>\models\checkpoints before running an image workflow. The installer
does not download the large SDXL Lightning model automatically.
Models, input, output, saved workflows, UI settings, and custom nodes live under
DataRoot. Rerunning with the same root and port leaves a healthy container in
place. To stop only this installation, use:
docker --context desktop-linux stop ods-comfyui-standalone
Rerunning the installer with the same data root and port restores the service. No installer command deletes the data root. This standalone path currently supports NVIDIA GPUs on Windows; it does not enable Dockerized ROCm on AMD.