manueljpconde makes Mike run on your own machine, with your own model

manueljpconde has added a single-command local setup that stands up the whole application - interface, backend, database, login, and file storage - on one machine, with no cloud API keys required to get it running. For a firm that wants to kick the tyres without signing up for hosted services or exposing client data, that's the difference between a demo and a non-starter.

The bigger piece is what the team calls managed models: each user can point Mike at their own AI engine - a local model running on their own hardware, or an account with Microsoft's Azure OpenAI service - and supply their own key, which is encrypted before it's stored and never handed back. It turns "which AI is reading my documents, and where does it live" into a setting the customer controls rather than a decision baked in upstream.