This is a user interface for your browser written in Python with the idea to provide a minimal UI for chat with local LLMs using llama.cpp featuring Retrieval Augmented Generation (RAG) and audio input using whisper.cpp.
The UI offers simple user identification without password authentication. This mainly serves to have a history with each user.
- Checkout or download zip.
python3 -m venv env
. env/bin/activate
pip install -r requirements.txt
- in the directory:
PYTHONPATH=<path>/server python3 server/app.py
- Independently start the llama.cpp server
./server -m ~/Downloads/models/dolphin-2.6-mixtral-8x7b.Q6_K.gguf --threads 8 -ngl 100 -c 32768 --cont-batching --parallel 1 -b 128
- Independently start the GROBID docker image and expose it
docker run --rm --init --ulimit core=0 -p 8070:8070 grobid/grobid:0.8.0
npm install
npm run build
npm run build_mini
- Get rid of langchain