isair/jarvis

A 100% private AI voice assistant that lives on your computer (works offline). Talk naturally as if Jarvis is a third person in the room, and get conversational responses. It remembers everything, knows location and time, can check the web, control Chrome, track nutrition, and more with support for unlimited MCPs / tools without context rot.

What it solves

Jarvis is a 100% private, offline AI voice assistant designed to act as a "third person in the room." It eliminates the need for cloud-based subscriptions and data harvesting by processing all AI tasks locally on the user's computer. It solves the problem of rigid, command-based voice assistants by providing conversational awareness, allowing it to understand context from ongoing discussions without requiring the wake word for every single turn.

How it works

Jarvis combines several local AI components to create a seamless voice interface:

  • Speech Recognition: Uses Whisper (with optional CUDA acceleration on Windows) to transcribe speech to text.
  • LLM Orchestration: Integrates with local LLM providers like Ollama or OpenAI-compatible servers (LM Studio, Jan, etc.) for chat and intent classification.
  • LLM Intent Judge: A small, fast model (e.g., gemma4:e2b) is used for real-time tasks like echo detection and query extraction.
  • Memory & Knowledge: Employs a self-organizing knowledge graph and embedding-based semantic search to maintain unlimited memory across conversations.
  • Tool Integration: Uses the Model Context Protocol (MCP) to connect to external tools (GitHub, Slack, Home Assistant) and an embedding-based router to select relevant tools without slowing down the system.
  • Text-to-Speech: Uses Piper TTS by default for fast, local voice synthesis.

Who it’s for

  • Privacy-conscious users who want a powerful AI assistant without sending their data to the cloud.
  • Power users who and want to integrate their AI assistant with their local files, smart home, and other developer tools via MCP.
  • Individuals seeking a high-quality, offline dictation tool as a free alternative to services like WisprFlow.

Highlights

  • Conversational Awareness: Can chime into discussions and understand context from a rolling temporary context window.
  • Unlimited Local Memory: Features a knowledge graph and a Memory Viewer GUI to track what the AI remembers.
  • Private Dictation Mode: A global hotkey allows users to paste transcribed speech into any application offline.
  • Extensible via MCP: Supports unlimited tools via the Model Context Protocol without performance degradation.
  • Automatic Redaction: Automatically redacts sensitive information before saving data to disk.
  • Adaptive Tone: Automatically adjusts its tone based on the topic (e.g., surgical for code, encouraging for wellbeing).

Related

  • Project
    LYiHub/pub-local-jarvis

    A local multimodal desktop assistant for Windows that perceives screen and system audio to provide real-time gaming companionship, course notes, and contextual AI interaction.

  • Project
    akshayaggarwal99/jarvis-ai-assistant

    An open-source voice dictation tool that converts speech to text anywhere on your system, supporting local transcription and LLM post-processing for privacy and speed.

  • Project
    AnubhavChaturvedi-GitHub/jarvis-ai-assistant

    A modular, voice-activated AI desktop assistant that combines speech recognition, LLM reasoning, image generation, and system automation to control your computer hands-free.

  • Project
    alondmnt/joplin-plugin-jarvis

    An AI note-taking assistant for Joplin that enables chatting with your notes, semantic search, and automated literature reviews using online or offline LLMs.

  • Project
    jasoncheng7115/jt-live-whisper

    A fully local AI toolkit for real-time speech transcription, translation, and meeting summarization that works across any system audio source without cloud APIs.