🪲RH应用@炮老师的小课堂

RunningHub isn't just a walled set of flagship models - it's an app marketplace where users publish workflows as "apps," each with its own parameter schema. DapaoRHAppNode is the generic runner for all of them: you paste an 🆔 应用ID, feed the app's inputs through a buffet of generic image/video/audio/text/number/boolean ports, and it submits, polls, and returns whatever the app produces. It's the "long tail" node of the RH section - the one that works even when the dedicated nodes haven't caught up to a model yet.

The inputs that matter

• 🌐 API渠道 + 🔑 API密钥 - the standing RH rule: domestic (runninghub.cn) and international (runninghub.ai) accounts use different keys; pick the channel that matches your registration.
• 🆔 应用ID - the RunningHub app identifier. This is the field everything hinges on; the app's own page documents what its inputs are.
• 🧩 应用参数JSON - a JSON object for app parameters that don't fit the generic ports.
• ⚙️ 实例类型 - three choices, default 默认. Some RH apps let you pick a compute/quality tier; this is the dial.
• The generic input buffet: 🖼️ 图像1–8, 🎞️ 视频1–4, 🎵 音频1–4, 📝 文本1–16, 🔢 数字1–8 (FLOAT), 🔘 布尔1–8 (BOOLEAN). Sixteen text slots is a lot of room for prompt-based apps; eight boolean slots covers the toggle-heavy ones.
• 📋 额外节点参数JSON (default []) - an array form for node-level params the app expects.
• Polling/robustness: 🔁 最大轮询秒数 (2400 - the longest in the RH set, for long-running apps), ⏱️ 轮询间隔 (5s), ⌛ 请求超时 (120s), 🚫 出错时跳过.

Outputs

🖼️ 图像 (IMAGE), 🎬 视频 (VIDEO), 🆔 任务ID, 🔗 输出URL, 📋 响应信息. You wire whichever the app produces; the others come back empty. The image and video outputs are tensors, so a video app's result drops straight into a save node.

How it works

Same async wrapper as the other RH nodes, generalized: submit to RunningHub's openapi/v2 app endpoint with the app ID and whatever generic inputs are wired, poll until done, then decode image/video results into tensors and pass URLs/status through. Because apps vary, the real work happens on RunningHub's side - this node is a pass-through with the plumbing done for you.

The honest read

This is the node to keep in your back pocket when the dedicated RH nodes (image, Seedance, V3.1, X-video3) don't cover a new model yet. RunningHub publishes new apps frequently; waiting for a pack update to get a dedicated node is slow, while this runner works the day the app appears. The cost is discoverability and typing: you need the app ID and you need to know (from the app's page) which inputs map where, because the generic ports won't tell you. And as always with this pack, the README's invite codes are the author's referral business - harmless, but know it's there.

Two things to keep honest:

• Wrong app ID / wrong input mapping fails quietly. The 响应信息 output is your debug channel; check it before assuming the app is broken.
• 2400s max polling tells you these apps can take a while. Longer than the dedicated video nodes, because arbitrary apps include slow pipelines. Build patience in.

Gotchas

• 国内版/国外版 key mismatch - the eternal RH trap, still the top failure mode.
• The generic ports map by position, not by name for some apps - an app expecting "prompt" as its second text input may need 文本2, not 文本1. Match against the app's docs.
• Inputs you don't need can still be sent if the app schema picks them up - wire only what the app declares.
• Chinese-only labels.

Install: ComfyUI Manager → "dapaoAPI", or git clone https://github.com/paolaoshi/ComfyUI-dapaoAPI.git → pip install -r requirements.txt → restart. RunningHub account + API key required.