manueljpconde teaches Mike to run on your own hardware

Until now, Mike's brain had to be one of three cloud services - Anthropic, Google, or OpenAI. manueljpconde adds a fourth path: point the system at a model running on your own server or a local setup (think Ollama, a popular tool for running AI models on your own machine), and client data never has to leave your walls. From a user's seat it's just another model name in the dropdown.

There are honest limits. Local models often can't reliably drive Mike's tool-using features, so when that's the case the fork quietly switches them off and tells the user why. There's also no automatic retry or fallback if the local model misbehaves, and the setup trusts whatever you declare about its capabilities rather than checking. This is a foundation, not a turnkey deployment.