可选模型面板
A live catalog of every model this pack can call — no plugin update needed
Every time the Wujiai channel adds a model, this panel is how you'd ever know. WujiaiModelBrowser (display name "可选模型面板", the "available models panel") is a read-only reference node that shows the current model catalog for the Wujiai API, split by category. It does nothing else - no generation, no uploads, no outputs at all. It's a poster on the wall, and that's the whole point.
The cloud-list trick
Here's the trick that makes it interesting. Most node packs hardcode their model dropdowns in Python, which means every new model requires a plugin update. This pack doesn't: the model list is a JSON file hosted in the cloud (a raw file on Gitee), fetched on demand. The browser node pulls that config, caches it locally in the pack's data/ directory, and re-fetches when it's stale - the README calls this "模型云端热更新" (cloud model hot-update). When Wujiai adds a new image or video model, the author just edits the cloud JSON and every installed copy of the pack sees it next run. That's also why so many model fields in this pack - like WujiaiSora2SubmitTask's model - are free text instead of dropdowns: the author can't enumerate choices that change weekly.
How the panel works
Mechanically it's a marriage of a Python node and a front-end script. The Python side registers an HTTP route on ComfyUI's server (/nkxx/wujiai/models_config) that returns the config, and the node itself exposes two widgets: category (a dropdown of 图像模型 / 视频模型 / 大语言模型 - image, video, LLM) and model_list_display (the text area that ends up showing the list). The bundled web/tools.js grabs the config from that route and fills the text area with the model names for whatever category you've selected, styled as a monospace, read-only terminal panel with a "顶部公告" (top announcement) header if the cloud file ships one. Switch category, the panel redraws. That's it.
Using it and installing
So how do you actually use it? As a lookup table while you build workflows. Open the panel, flip to 视频模型, and read off the exact model ID strings to type into the free-text model fields of the submit nodes. The node has no outputs and is flagged as an output node, so it can sit at the edge of your graph as a pinned reference without breaking anything. If the model you want isn't in the list, that's your answer: either it's not available on the Wujiai channel or you're looking at the wrong category.
Install is the pack routine - clone https://github.com/jieg9341-lab/ComfyUI-nkxx into ComfyUI/custom_nodes/, restart, let it auto-install the six deps. No API key needed here; it's public catalog data, not your account. The one real gotcha is network: it fetches from Gitee with a 5-second timeout, so if you're behind a firewall or Gitee is unreachable it silently falls back to the hardcoded defaults plus whatever's in the local cache. If the panel ever looks stale, clear the cached JSON in the pack's data/ folder and rerun - it'll refetch. And treat the announcement header as author news, not documentation; the actual model availability still lives with the API provider.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| model_list_display | STRING | 正在加载最新模型列表... | — |
| category | COMBO | 图像模型 | 3 options: 图像模型, 视频模型, 大语言模型 |
Outputs (0)
No outputs