Warning This repository is no longer maintained. Please see other forks, such as wfox4/WebChatRWKVv2. For more info on the RWKV language model, check the BlinkDL/ChatRWKV repository.
RWKV-V4 inference via rwkvstic, with a ChatGPT-like web UI, including real-time response streaming.
# Clone this repository
git clone https://github.com/hizkifw/WebChatRWKVstic.git
cd WebChatRWKVstic
# Recommended: set up a virtual environment
python -m venv venv
source ./venv/bin/activate
# Install requirements
pip install -r requirements.txt
# Run the webserver
python main.py
The script will automatically download a suitable RWKV model into the models
folder. If you already have a model, you can create the models
directory and
place your .pth
file there.
- Mobile-friendly web UI with autoscroll, response streaming, markdown formatting, and syntax highlighting
- Input is formatted into a question/answer format for the model, and earlier chat messages are included in the context
- Tune the model to better match ChatGPT
- Clean up the code
- Create a Docker image