Module Management

Module Management is the "look before you leap" node of the ComfyBros RunPod family. It calls a module_management workflow on your RunPod serverless endpoint and asks it to list or describe the models available there - then hands you the answer as a JSON string. Its own description spells out the job: "Manage models and modules using RunPod serverless instances via API calls. Supports listing and getting model information with optional filtering."

Why does this exist? Because the pack's generation nodes let you type a checkpoint name as a plain string, with zero validation. Type a model name that isn't on the endpoint and the job fails server-side, after you've already paid for the cold start. This node is the cheap sanity check: list what's installed, then wire the result into a text display or parse it with the pack's JsonParse/Dict Get nodes to confirm the exact filename before you tell Generate Image to use it.

Inputs

• instance_name - your RunPod endpoint, from the configured instances dropdown (same setup as the rest of the cluster; empty list means "No instances configured").
• management_action - list or get. list enumerates what's available; get pulls info about a specific model.
• The optional filters - model_kind, model_name, model_path, model_provider - are strings that narrow the query. Leave them empty and the node simply omits them from the payload.

Output

One output: result_json, a STRING of JSON. It wraps the endpoint's response with status, the action and params you sent, plus execution_time and delay_time - the pack likes its receipts. On a request failure you get an error JSON back with an error_type and message, so the node degrades to something you can still read rather than a bare traceback.

How it works

Same serverless contract as the other API nodes: POST {endpoint}/run with a workflow_name of module_management and your action/filters in workflow_params, then poll {endpoint}/status every couple of seconds until the job completes, with a 900-second timeout. Yes, even a model listing can sit in a cold-start queue - that's serverless for you.

Installing it

ComfyBros installs like any custom node pack. ComfyUI Manager - search "ComfyBros" - or:

cd ComfyUI/custom_nodes
git clone https://github.com/turnbros/ComfyBros

Then restart ComfyUI. The README's requirements.txt doesn't exist; the declared deps are in pyproject.toml. The node lives under ComfyBros/API Nodes/Model Management.

Gotchas

It only works if your endpoint actually implements a module_management workflow - this is a contract between the node and the RunPod worker, not a universal API. If the endpoint doesn't have that workflow, you'll get an error echoing the unexpected response. Configure an instance first, and remember this is a debugging companion: run it before generation, not alongside it, so you're not waiting on two cold starts at once.