manueljpconde lets mikeEU run on your own hardware

manueljpconde has wired mikeEU up to talk to local model runners like Ollama and LM Studio - tools that let you run an AI model on your own server instead of sending every conversation to a hosted provider. For firms with data-sensitivity concerns, that's the whole ballgame.

The plumbing is careful. The local endpoint's address and credentials stay on the backend; the user-facing app only learns that a local option exists and roughly what it can do. Function-calling - the AI's ability to invoke tools mid-conversation - is switched off by default, because many self-hosted models handle it unreliably. An operator has to explicitly turn it on, and until they do, the local provider quietly hides from the parts of the interface that would offer features it can't deliver.