elitan wires Mike up for fully local AI

elitan has added Ollama as a third option alongside the existing cloud AI providers. Ollama is a tool that runs open-source language models directly on a laptop or server - no outside API calls, no per-token billing, nothing leaving the box. Models tagged for Ollama show up in the picker as always-available, and the usual nag about missing API keys quietly steps aside.

The direction of travel is clear from the rest of the day's work in this fork: the maintainer is pulling out other cloud dependencies in parallel. Taken together, it's a push to make Mike something a firm or solo practitioner could stand up entirely behind their own firewall. The streaming and tool-use plumbing is hand-rolled, so anyone leaning hard on tool calls against a small local model should kick the tyres first.