page

local-voice-ai Reference

local-voice-ai Reference Implementation

Overview

Complete Docker-based local voice assistant stack.

GitHub: https://github.com/ShayneP/local-voice-ai

Stack Components

LiveKit: WebRTC signaling
Whisper (VoxBox): Speech-to-text
llama.cpp: Local LLM
Kokoro: Text-to-speech (we won't use this)
FAISS + Sentence Transformers: RAG/knowledge retrieval
Next.js: Frontend UI

System Requirements

Docker + Docker Compose
No GPU required (CPU-based models)
Recommended: 12GB+ RAM

Quick Start

git clone https://github.com/ShayneP/local-voice-ai
cd local-voice-ai
./test.sh
# Access at http://localhost:3000

What We Can Learn From This

Docker service orchestration pattern
LiveKit agent configuration
Whisper integration for STT
How to swap out TTS component

Our Modifications

Replace Kokoro TTS with InWorld AI (our existing voice pipeline)
Replace llama.cpp with Ollama running lars-trained
Add wake word detection layer (openWakeWord)
Add Claude Code SSH delegation

🌳 View Tree