Happy slice of mcp pizzaMCP.pizza

aurora

MCP.Pizza Chef: joaojhgs

Aurora is a privacy-focused intelligent voice assistant client that enables hands-free, voice-driven interaction with computers. It uses real-time speech-to-text conversion, large language models like Llama 3 and Mistral 7B, and integrates with open-source tools such as OpenRecall for semantic search and browser-use for automation. Aurora supports offline wakeword detection and local processing to ensure user data privacy while boosting productivity through seamless automation workflows.

Use This MCP client To

Activate voice assistant with custom wakeword offline Convert speech to text in real-time locally Perform semantic search on daily activities Automate browser tasks via voice commands Interact hands-free with computer applications Integrate local LLMs for natural language processing Enable privacy-focused voice-driven workflows Use open-source tools for automation and search

README

Aurora: Intelligent Voice Assistant for Local Automation and Productivity

Aurora is an intelligent voice assistant designed to enhance productivity through local, privacy-focused automation. It leverages real-time speech-to-text, a large language model (LLM), and open-source tools to provide a seamless and intuitive user experience. Aurora integrates with tools like OpenRecall for semantic search of daily activities and browser-use for browser automation, enabling users to interact with their computer in a hands-free, voice-driven manner.

Features

  1. Wakeword Detection:

    • Activate the assistant with a custom wakeword (e.g., "Jarvis").
    • Offline and low-latency detection using OpenWakeWord.

  2. Real-Time Speech-to-Text (STT):

    • Convert user speech into text using Whisper (OpenAI's lightweight model for local processing).

  3. Large Language Model (LLM) Integration:

    • Use Llama 3 or Mistral 7B (quantized for efficiency) to process user queries and generate responses.
    • Orchestrate tool calls (e.g., OpenRecall, browser-use) using LangChain and Langgraph.

  4. Semantic Search with OpenRecall:

    • Index and retrieve information from periodic screenshots and activities using OpenRecall.
    • Enable queries like, "What did I research about interfaces at 2 PM?"
    • Enrich the assistant context by adding past activities when necessary

  5. Browser Automation:

    • Control web browsers (e.g., open tabs, fill forms, click elements) using the browser-use framework.
    • The assistant will interpret your request, deem wether it should call the browser-use or not, and finally re-structure your request so that it's carried out correctly.

  6. Text-to-Speech (TTS):

    • Generate natural-sounding audio responses using Piper (offline TTS).

  7. Local and Privacy-Focused:

    • All processing happens locally, ensuring data privacy and security.
    • No cloud dependencies or data sharing.
    • Kinda, using OpenAI while development, but local LLMs are nativelly supported with langchain

  8. Modular Tooling and Integrations:

    • All integrations and tools are available through plugins which you can activate through the envs
    • Only install dependencies for the plugins you'll want to use, keeping the sizes low
    • Easy setup, just need to activate it and fill the correct env credentials if necessary

Libraries and Tools

aurora FAQ

How does Aurora ensure user privacy?
Aurora processes speech and commands locally using offline wakeword detection and local LLMs like Llama 3 and Mistral 7B, minimizing data sent to external servers and protecting user privacy. Q: Can Aurora work without an internet connection?