Say I wanted to run this on my phone today. What's the best path one could take to actually use it as mkre than a toy? I suspect step 1 is to get llama.cpp runn

Say I wanted to run this on my phone today. What's the best path one could take to actually use it as mkre than a toy? I suspect step 1 is to get llama.cpp running in Termux and a ggml or other model file loaded. That gets you a local web interface for inference. Then... how do you hook that up to your system such that it can do anything useful?