MARK XL is a cross-platform, local-first AI assistant designed for privacy-conscious users seeking OS-level control without relying on cloud APIs. It integrates offline speech recognition, a locally hosted LLM via Ollama, and text-to-speech capabilities to provide a responsive and secure personal assistant experience.

How It Works

MARK XL employs a modular architecture: Microphone input is processed by offline Speech-to-Text (STT) engines like Whisper or Vosk. The resulting text is fed to a locally hosted LLM (managed by Ollama), which handles tool calling and streaming responses. Executed tools interact with the OS, browser, or files. Finally, Text-to-Speech (TTS) engines such as EdgeTTS, Kokoro, or ElevenLabs convert the LLM's output back into speech. This local-first approach prioritizes user privacy and enables real-time interaction and control.

Quick Start & Requirements

1. Install Ollama: Follow instructions at https://ollama.com.
2. Pull an LLM model: Execute ollama pull qwen2.5:7b (or another desired model).
3. Clone/download the project: Obtain the Mark-XLVI repository.
4. Launch: Navigate to the project directory and run python main.py.

Requirements: Python 3.11 or 3.12, Ollama installed with at least one model pulled, and a functional microphone. The initial run includes automatic installation of base Python packages and subsequent engine setup.

Highlighted Details

• Streaming Responses: TTS begins speaking immediately upon receiving the first sentence, enhancing perceived responsiveness.
• Extensive Tooling: Features 18 built-in tools for OS control, browser automation, file management, messaging, web searches, and more.
• Live Configuration: Allows dynamic changes to STT, LLM, and TTS engines, voices, and models via a UI without restarting the application.
• Long-Term Memory: Persistently stores personal facts and recalls them across conversations.
• Cross-Platform: Supports Windows, macOS, and Linux environments.
• Offline Operation: Core components (STT, LLM via Ollama, Kokoro TTS) function entirely offline, ensuring data privacy.

Maintenance & Community

Information regarding specific maintainers, community channels (e.g., Discord, Slack), or project roadmaps was not detailed in the provided README.

Licensing & Compatibility

The project is released under the MIT License. This permissive license generally allows for commercial use and integration into closed-source projects.

Limitations & Caveats

Performance is dependent on local hardware capabilities for running STT, LLM, and TTS models. While core functionality is offline, the default EdgeTTS engine requires an internet connection; the fully offline Kokoro TTS is an alternative. ElevenLabs TTS is a paid service. The initial setup involves background package installations and engine configuration.