comfyui
| Version | 2026.144.1531 |
| Repo | superproject |
GPU image-generation server with ComfyUI + the Manager and model-download tooling baked in
Clones ComfyUI to ~/ComfyUI plus the ComfyUI-Manager custom node and
the full model directory tree, and installs the aria2 + git-lfs tooling
used to fetch model checkpoints. A supervisord service runs
main.py --listen 0.0.0.0 on port 8188. Each piece is verifiable: the
cloned entrypoint and Manager, the model dirs, the download-tooling
packages, and the live /system_stats API on port 8188.
Services
Section titled “Services”comfyui
Acceptance plan
Section titled “Acceptance plan”This candy’s plan: — the runnable spec charly check executes against a live deployment. check: steps are idempotent probes; run: steps change state.
| Intent | Step |
|---|---|
run |
command=git clone –depth 1 https://github.com/comfyanonymous/ComfyUI.git ${HOME}/ComfyUI || true cd ${HOME}/ComfyUI mkdir -p models/{checkpoints,clip,clip_vision,configs,controlnet} mkdir -p models/{diffusers,diffusion_models,embeddings,gligen,hypernetworks,loras} mkdir -p models/{photomaker,style_models,text_encoders,unet,upscale_models,vae,vae_approx} mkdir -p custom_nodes output input user git clone –depth 1 https://github.com/ltdrdata/ComfyUI-Manager.git ${HOME}/ComfyUI/custom_nodes/ComfyUI-Manager || true |
check |
the ComfyUI server entrypoint main.py is baked into ~/ComfyUI |
check |
the ComfyUI-Manager custom node directory is cloned alongside it |
check |
aria2c is installed so model checkpoints can be fetched in parallel |
check |
git-lfs is installed for pulling large model repositories |
agent-check |
the comfyui supervisord service comes up and serves the image-generation web UI and /system_stats API on port 8188 |
check |
file=${HOME}/ComfyUI/models/checkpoints |
check |
package=aria2 |
check |
file=/usr/bin/git-lfs |
check |
service=comfyui |
check |
addr=127.0.0.1:${HOST_PORT:8188} |
check |
http=http://127.0.0.1:${HOST_PORT:8188}/system_stats |