Dobby Voice Assistant

Dobby is a home voice assistant that combines local speech processing and device control using a local LLM for inference and a remote LLM for broader requests. The bulk of functionality runs on an Ubuntu Linux i7 mini PC. An ESP32 audio frontend listens for a wake word, the Linux host interprets the spoken request, and WiFi networked speakers play the reply. The system also controls MQTT devices, plays a Jellyfin music library, and speaks notifications from other home services.

The aim is to make everyday home interactions convenient and personal: common commands stay on the local network, more open-ended requests can use an agent with tools, and responses use a consistent Dobby-style voice and "house elf" personality.

This is an evolving, installation-specific project. The github repository README describes the source and configuration reviewed in September 2026; it is not yet a tested clean-install guide. Supporting services, model assets, and firmware live partly outside the github project folder. See setup, operations, and the project review there.