Workflow Doctor

Workflow Doctor

Drop in a ComfyUI workflow → get the exact list of custom nodes and model files your install is missing.

No more "red nodes, missing custom node, required model not found" after downloading someone else's workflow. Workflow Doctor checks the workflow against your ComfyUI install and hands back a clean shopping list.

What it does

• Missing custom nodes — every node class in the workflow that your install cannot execute, with the node type name to search in ComfyUI-Manager.
• Missing model files — every .safetensors / .ckpt / .gguf / .pt referenced by the workflow that is not on your disk, with one-click searches on HuggingFace and CivitAI.
• Works on the canvas or on a file — check what you are about to queue, or check a workflow.json you just downloaded, before you open it.
• Also usable as a node — the Workflow Doctor node takes workflow JSON and returns a markdown report you can wire into anything.

Install

Via ComfyUI-Manager: search for Workflow Doctor in Install Custom Nodes → install → restart ComfyUI.

Manually:

cd ComfyUI/custom_nodes
git clone https://github.com/lostmywind/comfyui-workflow-doctor

Restart ComfyUI. No dependencies, nothing is downloaded, no account, no telemetry — it only reads the workflow and compares it against files already on your machine.

How to use it

1. Look for Workflow Doctor in the sidebar (or the 🩺 button in the bottom-right corner on older front-ends).
2. Hit Check canvas workflow for the graph you are looking at, or Open a workflow file… for a file on disk.
3. Read the report: what is missing, and where to get it.

Right-clicking the canvas also gives you Check workflow with Doctor.

Honest limitations

• It reports what is missing; it does not download model files for you (they are large and often license-gated — you pick the source).
• Node → repository mapping is not attempted. ComfyUI-Manager's own search handles installation, and guessing repos produces wrong answers.
• A model file that was renamed locally will show as missing, because matching is by filename.

Development

The detection engine (core.py) is plain Python with no ComfyUI imports — it receives the local environment as input, so it can be unit-tested offline:

python3 test_core.py

License

MIT — see LICENSE.