A Mac-based local AI assistant linked to a small hardware board

A personal named Belochka, also called Bella, runs on a Mac and connects to an ESP32 W-10 board used like a small dedicated device. LM Studio runs a model from a GGUF file in the background.

The assistant can send files, generated music, and other output from the Mac through a . It listens continuously, stays quiet when it has nothing useful to add, and can join in when needed.

The setup is built from Mi, a from , and a locally running open-source Xiaozhi server. It also uses or builds on tools and models such as llama.cpp, LM Studio, Xiaozhi, , Mi, Anthropic Fable, and Qwen AI Lab, with no stated sponsorship.

Key points

  • The assistant on a Mac and connects to an ESP32 W-10 board.
  • LM Studio runs a model from a GGUF file.
  • A lets the assistant send files and generated music from the Mac.
  • The flow is Mi → → local Xiaozhi server.
  • An always-listening assistant needs careful privacy and security settings.
Read original