how to work with llama.cpp

I'm on a Mac (M1) and generally use llama.cpp with model, e.g., llama 3 instruct 8B....gguf.  I want to use Piper, which I've used before and it's fast enough that it makes up for its reduced audio quality (for now!).  To use AlwaysReddy, I think I have to edit config.py to set COMPLETIONS_API and COMPETION_MODEL to get this to work, right?  Any help with what I need to put here?  Thanks.