How do you improve integration when working with llama.cpp?

[owlchloe]

I have a question about llama.cpp that may be useful for other people working with it too.

I am particularly interested in integration and accessibility. I would like to compare approaches that people have actually used rather than just theoretical suggestions.

Which approach would you recommend and why?