elitan teaches Mike to run its AI on your own hardware

elitan wired in Ollama, a tool for running AI models locally on your own machine, as a fourth engine alongside the cloud heavyweights the fork already speaks to (Anthropic's Claude, Google's Gemini, and OpenAI). It ships switched off by default; an operator turns it on with a single configuration flag, after which the local models show up in the picker.

The interface side is just as plain. When a local engine is available, the models page grows a one-click button to swap whatever cloud model you're on for a local equivalent. Nothing about the existing cloud paths changes - this is a bolt-on alternative, not a rewrite. The set of local models on offer is deliberately small, so any team adopting it would pick its own.