A newer version of the Gradio SDK is available:
4.42.0
川虎 Chat 🐯 Chuanhu Chat
Lightweight and User-friendly Web-UI for LLMs including ChatGPT/ChatGLM/LLaMA
Compatible with GPT-4 · Chat with files · LLMs local deployment · Web search · Chuanhu Agent · Fine-tuning
Video Tutorial · 2.0 Introduction · 3.0 Introduction & Tutorial || Online trial · One-Click deployment✨ 5.0 Major Update!
New! An all-new user interface! So exquisite that it doesn't look like Gradio, it even has a frosted glass effect!
New! Adapted for mobile devices (including perforated/bezel-less phones), the hierarchy is clearer.
New! The history is moved to the left for easier use. And supports search (with regular expressions), delete, and rename.
New! Now you can let the large model automatically name the history (Enabled in the settings or configuration file).
New! Chuanhu Chat can now be installed as a PWA application for a more native experience! Supported on Chrome/Edge/Safari etc.
New! Icons adapted for all platforms, looking more comfortable.
New! Supports Finetune (fine-tuning) GPT 3.5!
Supported Models
API Callable Models | Remarks | Locally Deployed Models | Remarks |
---|---|---|---|
ChatGPT(GPT-4) | Support fine-tune gpt-3.5 | ChatGLM (ChatGLM2) | |
Azure OpenAI | LLaMA | Support Lora models | |
Google PaLM | Not support streaming | StableLM | |
iFlytek Starfire Cognition Large Model | MOSS | ||
Inspur Yuan 1.0 | Qwen | ||
MiniMax | |||
XMChat | Not support streaming | ||
Midjourney | Not support streaming | ||
Claude |
Usage Tips
💪 Powerful Functions
- Chuanhu Assistant: Similar to AutoGPT, automatically solves your problems;
- Online Search: Is ChatGPT's data too old? Give LLM the wings of the internet;
- Knowledge Base: Let ChatGPT help you speed read quantumly! Answer questions based on files.
- Local LLM Deployment: One-click deployment, get your own large language model.
🤖 System Prompt
- The system prompt can effectively enable role-playing by setting prerequisite conditions;
- ChuanhuChat presets Prompt templates, click
Load Prompt Template
, choose the Prompt template collection first, then choose the Prompt you want in the list below.
💬 Basic Conversation
- If the answer is not satisfactory, you can try the
Regenerate
button again, or directlyDelete this round of conversation
; - Input box supports line breaks, press Shift + Enter to make one;
- Using the ↑ ↓ arrow keys in the input box, you can quickly switch between send records;
- Generating a new conversation every time is too cumbersome, try the
single-dialogue
function; - The small button next to the answer bubble not only allows
one-click copy
, but also lets youview the original Markdown text
; - Specify the answer language, so that ChatGPT will always reply in a certain language.
📜 Chat History
- Dialogue history will be automatically saved, you won't have to worry about not being able to find it after asking;
- Multi-user history isolation, only you can see them;
- Rename chat, easy to find in the future;
- New! Magically auto-name the chat, let LLM understand the conversation content, and automatically name the chat for you!
- New! Search chat, supports regular expressions!
🖼️ Small and Beautiful Experience
- Self-developed Small-and-Beautiful theme, gives you a small and beautiful experience;
- Automatic light and dark color switching, gives you a comfortable experience from morning till night;
- Perfectly rendering LaTeX / tables / code blocks, supports code highlighting;
- New! Non-linear animations, frosted glass effect, so exquisite it doesn't look like Gradio!
- New! Adapted for Windows / macOS / Linux / iOS / Android, from icon to screen adaptation, gives you the most suitable experience!
- New! Supports PWA app installation for an even more native experience!
👨💻 Geek Functions
- New! Supports Fine-tuning gpt-3.5!
- Plenty of available LLM parameters to adjust;
- Supports API-host switching;
- Supports custom proxies;
- Supports multiple api-key load balancing.
⚒️ Deployment Related
- Deployment to the server: Set in
config.json
"server_name": "0.0.0.0", "server_port": <your port number>,
. - Obtain public link: Set in
config.json
"share": true,
. Note that the program must be running to access it through public links. - Use on Hugging Face: It's recommended to Duplicate the Space in the top right corner before using, the App response might be faster.
Quick Start
Execute the following commands in the terminal:
git clone https://github.com/GaiZhenbiao/ChuanhuChatGPT.git
cd ChuanhuChatGPT
pip install -r requirements.txt
Then make a copy of config_example.json
, rename it to config.json
, and then fill in your API-Key and other settings in the file.
python ChuanhuChatbot.py
A browser window will automatically open, at this point you can use Chuanhu Chat to chat with ChatGPT or other models.
Note
Please check our wiki page for detailed instructions.).
Troubleshooting
When you encounter problems, you should try to manually pull the latest changes1 and update dependencies2 first, then retry. Steps are:
- Click on the
Download ZIP
button on the website, download the latest code and unzip to replace, orgit pull https://github.com/GaiZhenbiao/ChuanhuChatGPT.git main -f
- Try to install dependencies again (the project might have new dependencies)
pip install -r requirements.txt
Generally, you can solve most problems by following these steps.
If the problem still exists, please refer to this page: Frequently Asked Questions (FAQ)
This page lists almost all the possible problems and solutions. Please read it carefully.
More Information
More information could be found in our wiki:
- How to contribute a translation
- How to make a contribution
- How to cite the project
- Project changelog
- Project license
Starchart
Contributors
Sponsor
🐯 If you find this project helpful, feel free to buy me a coke or a cup of coffee~