ChuanhuChatGPT (川虎 Chat) is a free, GPL-3.0 Python and Gradio web interface that puts the ChatGPT API and more than a dozen other hosted and locally deployed large language models behind one self-hostable chat UI, built for individuals and small teams who want their own front end instead of a vendor's.

What it is

ChuanhuChatGPT is a web-based graphical interface for large language models, written in Python and distributed under GPL-3.0. It runs as a local web application: the user supplies API keys or points it at a locally served model, and the app presents every supported model through one chat window. The supported set spans hosted APIs such as ChatGPT (GPT-5, GPT-4, GPT-4o, o1), Azure OpenAI, Google Gemini Pro, Claude, DeepSeek, MiniMax, 讯飞星火, Inspur Yuan 1.0, Midjourney, DALL·E 3 and XMChat, alongside locally deployed models including ChatGLM, ChatGLM2, ChatGLM3, LLaMA (with Lora models), StableLM, MOSS and Qwen (通义千问). The README is published in 简体中文, English, 日本語, Russian and 한국어, and a hosted instance exists as a Hugging Face Space.

The concrete problem it solves is fragmentation. Each provider ships its own console, its own key management and its own API quirks, and a user who wants file-based question answering, web search, an agent, or fine-tuning normally assembles that from separate scripts. ChuanhuChatGPT replaces those per-vendor front ends with one application driven by a single configuration file, so switching models does not mean switching tools. It also extends past plain chat by bundling retrieval over local files, live search, agent execution, custom model endpoints and GPT-3.5 fine-tuning into the same interface.

Key capabilities

• Unified model selector covering hosted APIs and local deployments, including custom model definitions that point at a local inference service.
• 川虎助理, an agent that works through a problem automatically in the manner of AutoGPT.
• 联网搜索, which lets a model query the web instead of relying on stale training data.
• 知识库, a file-based question-answering mode that answers from documents the user supplies.
• GPT-3.5 fine-tuning, plus adjustable LLM parameters, custom api-host, custom proxy and multi-api-key load balancing.
• A history system that saves conversations automatically, isolates them per user, supports renaming, automatic LLM-generated naming and regex search.
• Rendering for LaTeX, tables and highlighted code blocks, a Small-and-Beautiful theme with automatic light/dark switching, and installation as a PWA on Chrome, Edge or Safari.

Who uses it and how

• Single developers running it locally on Windows, macOS, Linux, iOS or Android by cloning the repository and starting the Python script.
• Small teams deploying it to a server by setting server_name to 0.0.0.0 and a server_port in config.json, where per-user history isolation matters.
• Users without suitable hardware trying the hosted Hugging Face Space, with the README recommending the Space be copied before use for faster response.
• Users who need remote access by setting share to true in config.json to obtain a public link, which stays valid only while the program keeps running.
• Chinese-speaking and other non-English users, who get prompts, templates and documentation in their own language.

Getting started

Clone the repository, run pip install -r requirements.txt, copy config_example.json to config.json and fill in the API key, then run python ChuanhuChatbot.py, which opens a browser window. The hosted Hugging Face Space is an alternative, and the project wiki covers detailed installation.

How it compares

No comparable tools or paid alternatives are named in the supplied facts, so this project stands alone in this registry.

When to use it — and when not to

A self-hoster must operate a Python environment and keep configuration, API keys, and any proxy or custom endpoint working; there is no managed hosting beyond the shared Hugging Face Space. The licence is GPL-3.0, so anyone distributing a modified version carries copyleft obligations. Two supported backends, XMChat and Midjourney, do not support streaming, and with 129 open issues the project carries a normal maintenance backlog; it is a poor fit for anyone wanting a zero-configuration commercial product or guaranteed streaming on every model.

git clone https://github.com/GaiZhenbiao/ChuanhuChatGPT.git
cd ChuanhuChatGPT
pip install -r requirements.txt

python ChuanhuChatbot.py