ComfyUI-LLMs

A minimalist node that calls LLMs, combined with one API, can call all language models, including local models. 一个极简的调用LLMs的节点，结合one-api，可以调用所有语言大模型，包括本地模型。

Supports gemini, glm-4-v, and qwen-v visual models, which can be used to infer cue words. 支持gemini、glm-4-v、qwen-v视觉模型，可以用来反推提示词。

Combined with one API, it supports almost all major language models. 结合one-api，几乎支持所有大语言模型。

note

To reduce the usage of tokens, by default, the seed remains fixed after each generation. Without changing the prompt words, clicking on generate will not trigger a response. If you want to generate a new response, you need to change the prompt words.

Installation

Git clone this repo into the custom_nodes folder. If necessary, check the pip requirements. It will be necessary.

Settings

The settings are stored in the settings.yaml file.

one-api settings: https://github.com/songquanpeng/one-api

Reference code

Based on the secondary development of the following project, some fields still use the original project. As a designer, writing code is not easy. If it's helpful, you can contribute together or add a star! cyberdolphin：https://github.com/whatbirdisthat/cyberdolphin ComfyUI_GLM4Node：https://github.com/JcandZero/ComfyUI_GLM4Node ComfyUI-Qwen-VL-API：https://github.com/ZHO-ZHO-ZHO/ComfyUI-Qwen-VL-API

License

GPL 3.