Chuanhu Chat (ChuanhuChatGPT) is an open-source project that provides a lightweight, easy-to-use web graphical interface and many additional features for ChatGPT and various other LLMs. According to its GitHub repository information, it supports models such as GPT-5 (including GPT-5, GPT-5-mini, GPT-5-nano; 400k context, up to 128k output) and DeepSeek R1.
Main capabilities include: file-based Q&A, local LLM deployment, web search, Agent assistants, and support for Fine-tune (fine-tuning GPT 3.5). The Chuanhu assistant is similar to AutoGPT and can solve problems fully automatically; online search can supplement network information for the LLM; the knowledge base can answer questions based on files; local LLM deployment supports one-click deployment; GPT 3.5 fine-tuning makes ChatGPT more personalized; it also supports custom models, such as connecting to local inference services.
How to use: In the project folder, copy config_example.json and rename it to config.json, then fill in settings such as the API-Key. After running, the browser window will open automatically, and you can use ChuanhuChat to converse with ChatGPT or other models. To deploy to a server, you can set "server_name": "0.0.0.0", "server_port": <port number> in config.json; to obtain a public link, you can set "share": true, but note that the program must be running for the public link to be accessible. It can also be used on Hugging Face, and it is recommended to duplicate the Space before using it.
Interface features include: a brand-new user interface with a frosted glass effect; adaptation to mobile devices (including the punch-hole/notch on full-screen phones); history moved to the left side, with support for search (including regex), deletion, and renaming; the large model can automatically name history records; it can be installed as a PWA application, with support for browsers such as Chrome/Edge/Safari; a self-developed Small-and-Beautiful theme with automatic light/dark mode switching; perfect rendering of LaTeX/tables/code blocks, with code highlighting support; and support for multi api-key load balancing.
Other features: role-playing by setting preconditions through System Prompt, with preset Prompt templates; if you are not satisfied with an answer, you can regenerate it or delete this round of conversation; the input box supports line breaks (Shift + Enter); pressing the ↑ ↓ arrow keys allows quick switching among send history; single-round conversation functionality; buttons next to answer bubbles allow one-click copying and viewing the original Markdown; you can specify the answer language; conversation history is automatically saved, with multi-user history isolation.
For installation and usage tutorials, please see the project's wiki page. If you encounter problems, you can first try manually pulling the latest changes and updating the dependency libraries: click Download ZIP to download the latest code and extract it to overwrite, or use git pull https://github.com/GaiZhenbiao/ChuanhuChatGPT.git main -f; then pip install -r requirements.txt. The FAQ page lists almost all possible problems you may encounter, including how to configure a proxy, etc.
