--prefer-binary
--extra-index-url https://abetlen.github.io/llama-cpp-python/whl/cpu
llama-cpp-python==0.3.32
protobuf<4
click<8.1
huggingface_hub
fastapi
uvicorn[standard]