Pick a back-up on-device model

* llama-70b runs locally but is very slow
* llama-7b is faster but can't follow the instructions
* I should try llama-13b, Gemma, or any other model supported by ollama.

The issue names no files, tests, or entry points. Start by locating the Ollama model-selection entry point and checking how the current llama-70b and llama-7b options are configured. Evaluate llama-13b, Gemma, and other supported models for speed and instruction following; done means selecting a suitable backup model and recording the choice.

Written by the indexing model from the issue text.