10 Workflow Documentation Agent

If you've ever shared a workflow and then had to explain it, you've felt the pain this node exists to automate. Workflow Documentation Agent takes a workflow name and your own summary/notes, and asks Gemini to write proper Markdown documentation for it - overview, requirements, node setup, usage steps, troubleshooting, release notes - using the section list you choose. If save_to_collection is on, the finished doc gets stored in your document collection, so you build a little documentation archive as you go.

"Share a workflow JSON, and the community expects the writeup too" is real ComfyUI culture (the KB's ecosystem essay covers how workflows became the currency), and this node is the pack's attempt to make the writeup less of a chore. It's a documentation generator, not a magic oracle: feed it a decent workflow_summary and you'll get a decent first draft.

How it works

The node builds a prompt from workflow_name, your workflow_summary or raw notes, the requested include_sections_csv, and any context_pack. Then it calls Gemini through the hosted ZMongo backend's /api/chat (via your text_agent_config, at a restrained temperature of 0.3). Notably, its own prompt instructs the model to explain "what the hosted backend does and why an API account is needed" and to include troubleshooting for auth/session/API-key failures - the author knows exactly where users get stuck.

The returned text is your markdown output. If save_to_collection is on and the generation succeeded, it's also written into collection_override (default text_agent_documents) as a workflow_documentation record, and you get the document_id back.

Inputs and outputs that matter

• text_agent_config - from a Text Agent Session node. Required; hosted API session needed.
• workflow_name - what to title the doc.
• workflow_summary - your notes about what the graph does. The quality of the output tracks this input directly.
• include_sections_csv - default overview,requirements,node setup,usage steps,troubleshooting,release notes. Trim it if you want a leaner doc.
• save_to_collection / collection_override - whether to persist the doc and where.

Outputs: markdown (STRING - the doc, ready to paste or save), json, document_id, and success.

Installing it

cd ComfyUI/custom_nodes
git clone https://github.com/CentralFloridaAttorney/ComfyUI-ZMongo

Restart ComfyUI, or install via ComfyUI Manager ("ComfyUI-ZMongo"). The pack's requirements.txt is heavy regardless - pymongo, langchain, sentence-transformers, transformers - so the first install is slow.

Common issues

• Empty markdown with success false - the Gemini call failed; check json for the raw backend error. Usually session/auth on the hosted side.
• success false but markdown exists - generation worked but the save to collection failed, and the node reports the combined result. Turn off save_to_collection if you only want the text.
• Docs read like marketing copy - the model is working from your workflow_summary. Give it concrete node names and data flow; it can only document what you told it.
• "Missing ZMongo API session" - Text Agents family needs the hosted API key session, not the local file store.