What is your app doing? Just LLM inference?

It's a custom agent harness with on-device models and the ability to swap between models.

Basically, a "toy" app to showcase where we are with coding agents on-device.