I looked into using https://github.com/mybigday/llama.rn. Ultimately, it was too slow to be conversational. The demands of the rendering the WebGL would likely not help.
FWIW, I've been running the dev builds for ages now (currently on 2.9.2015111), and it wasn't a big deal to update my (few) scripts. The new AppleScript dictionary is definitely a bit better, and simplified my scripts.
It was a while ago. If I was to do it over again I might try https://github.com/tirthajyoti-ghosh/expo-llm-mediapipe. Maybe newer models will help.