Browser LLM a bit weak I'm guessing.

Seems fine so far, I'm amazed actually. I'm just implementing controls for some of the underlying LLM parameters, and testing the results. It affects the way the model outputs, that's all.