Voice-powered terminal agent. Fully offline.
Press a key, speak, get an answer. Nothing leaves your machine.
| Meta |
|
| Powered by |
|
npm install -g voxagent
voxagent- Node.js 18+
- Ollama running locally
That's it. No API keys. No cloud accounts. No recurring costs.
On first run, voxagent downloads a small whisper model (~150 MB) for speech-to-text. Everything runs on your machine.
$ voxagent
Press ENTER to speak...
[Recording...] Press ENTER to stop.
Transcribing...
You: What's the default port for PostgreSQL?
Thinking...
PostgreSQL runs on port 5432 by default.
Press ENTER to speak...
--model <name> Ollama model to use (default: llama3.2)
--help, -h Show help
--version, -v Show version
voxagent captures your voice with decibri, transcribes it locally with whisper.cpp, sends the text to your local Ollama model, and prints the response.
No audio is recorded, stored, or transmitted. Ever.
- decibri - cross-platform microphone capture
- whisper.cpp - local speech-to-text
- Ollama - local LLM inference
Apache 2.0