Show HN: Offline voice assistant, NO dedicated GPU, less 15Gb RAM

Category: ai-ml

Tags: voice-assistant, offline, barge-in, tool-calling, local-llm

Score: 7.3/10 (Innovation: 7, Technical: 8, Documentation: 6, Utility: 7)

A fully offline voice assistant with barge-in, interruptible playback, and a layered turn-ending detector, running without a dedicated GPU under 15GB RAM. It features a tool gate for safe tool calling, two interchangeable reasoning backends, and an interrupt-safe conversation history, all coordinated by a hub process.

Target audience: backend devs, AI engineers, hobbyists

Repository: https://grigio.org/offline-voice-assistant-no-dedicated-gpu-less-15gb-ram/

View on Hacker News