Llama Assistant logo

Llama Assistant

Privacy-first AI desktop assistant

AI AssistantNewOpen Source

Completely offline AI assistant powered by Llama 3.2 and DeepSeek R1 with voice recognition, multimodal support, and RAG capabilities.

Llama Assistant

About Llama Assistant

Llama Assistant redefines personal AI assistance by providing a completely offline, privacy-first solution that rivals cloud-based alternatives. Powered by cutting-edge open-source models including Llama 3.2 and DeepSeek R1, it offers sophisticated AI capabilities without compromising your privacy.

The assistant features advanced voice recognition with wake word detection ("Hey Llama!"), allowing for natural voice interactions. Its multimodal capabilities enable it to understand and process both text and images, while the integrated RAG (Retrieval-Augmented Generation) system allows it to learn from your personal documents and knowledge base.

Built with privacy as a fundamental principle, Llama Assistant ensures that all your data, conversations, and personal information remain completely local to your device. This makes it perfect for professionals, researchers, and privacy-conscious users who want AI assistance without data concerns.

Key Features

Complete offline operation - no internet required
Voice wake word detection ("Hey Llama!")
Multimodal AI interactions (text + vision)
Screen capture OCR for context awareness
RAG knowledge base for document chat

Technologies

PythonLlama 3.2DeepSeek R1Speech RecognitionOCRRAGComputer Vision

Installation

Clone from GitHub:

git clone https://github.com/nrl-ai/llama-assistant.git