[simonw]

There's a Python binding for llama.cpp which is actively maintained and has worked well for me: https://github.com/abetlen/llama-cpp-python