Welcome to bitChat — your friendly, minimal, and modern AI chat demo powered by Microsoft Research’s BitNet! This project is a real-life implementation of the BitNet paper, showing off how you can chat with a cutting-edge 1-bit Transformer model in a ChatGPT-style UI. 🚀
- 🧑💻 Minimal, modern chat UI (React + Vite)
- 🤖 Real-time chat with BitNet (Microsoft Research)
- ⚡ FastAPI backend for model inference
- 🤝 Hugging Face Transformers integration
- 🐳 Dockerized for easy deployment (Azure-ready!)
- 🧩 Model selection UI (extensible)
- 🐳 Docker
- 🟢 Node.js (for local frontend dev)
- 🐍 Python 3.11+ (for local backend dev)
- Clone the repo:
git clone https://github.com/your-org/bitChat.git cd bitChat - Frontend:
cd frontend npm install npm run dev - Backend:
pip install -r requirements.txt uvicorn main:app --reload
- Open http://localhost:5173 and start chatting!
docker build -t bitchat:latest .
docker run -p 8000:8000 bitchat:latest- Build and push the image to Azure Container Registry
- Deploy using Azure Portal or CLI
- BitNet Paper: BitNet: Training 1-bit Transformers for Large Language Models
- Model: microsoft/bitnet-b1.58-2B-4T on Hugging Face
bitChat/
├── Dockerfile
├── main.py # FastAPI backend
├── inference.py # Model inference logic
├── requirements.txt
├── frontend/ # React + Vite frontend
└── static/ # Built frontend assets (served by backend)
This project is MIT licensed. See LICENSE for details.
🧪 This is a demo project for research and educational fun. Not production-ready!