🌐 RH Config

RH Config is the front door of the entire ComfyUI_RH_API pack. It doesn't generate anything, upload anything, or even call the network - it just packages your RunningHub credentials and target workflow into a single config object that every other node in the pack hangs off of. Think of it as the "login" node for a cloud GPU rental that lives outside your machine.

The background matters here: RunningHub is a cloud ComfyUI platform - the turnkey end of the "no local GPU" market. You pick a workflow, it's already wired up with models attached, and one click runs it on their hardware. This pack is what lets your local ComfyUI be that one click. It's the same "the model isn't on my machine" pattern as Runware or fal (covered in the KB's external-api-nodes essay), just pointed at RunningHub's API instead of a Western vendor.

What you actually set

Three required fields, and only two of them are mysterious:

• api_key - your RunningHub API key, from your account settings.
• workflow_or_app_id - the workflow ID you copy from "Run with API" on any workflow, or the numeric AI App ID if you're calling an app.
• base_url - leave this as https://www.runninghub.cn unless you're pointed at a mirror.

The optional is_ai_app checkbox switches the pack from the workflow endpoint to the AI App endpoint. If you set it wrong, you'll get a confusing task-creation error, so flip it only when you're genuinely calling an AI App rather than a workflow.

How it works under the hood

The node just builds a dict - api_key, workflow_or_app_id, base_url, is_ai_app - and hands it out as config. The one clever bit: if you leave api_key or base_url empty, it falls back to a config.json file sitting in the pack's directory. Copy config.json.example to config.json and drop your key in there, and you never have to paste it into a node again. It also prints where each value came from to the console, which is a nice touch when you're debugging "why is this using a weird key."

Wiring it up

The single config output goes to the config input of RH_Execute, RH_Download, RH_TaskManager, or any of the RH_Upload* nodes. One Config node can feed an entire graph - you don't need one per node.

Install

Easiest is ComfyUI Manager: search ComfyUI_RH_API, install, restart. By hand:

cd ComfyUI/custom_nodes
git clone https://github.com/xuchenxu168/ComfyUI_RH_API.git
cd ComfyUI_RH_API
pip install -r requirements.txt

Then restart ComfyUI. There are no model downloads - the models live on RunningHub's side, which is the whole point. (One oddity: the README's manual-clone URL says Ken-Chen-CN/..., but the current repo is xuchenxu168/... - either resolves to the same pack.)

Gotchas

The "API key is required" error is this node refusing to proceed because both the node field and config.json are empty. The "Workflow ID is required" one is the same for the ID. If you're sharing a workflow file with the key pasted in, remember you're sharing a live credential - that's the category's classic footgun (a malicious API node already got prosecuted once in this ecosystem, and your key is the prize). RunningHub itself also blocks NSFW content, so a workflow that depends on it will fail cloud-side no matter what you type here.