FatihMakes/Mark-LI

What it solves

Mark LI is a cross-platform personal AI assistant designed to act as a digital extension of the user, providing a hands-free way to control a computer and manage daily tasks through real-time voice and visual interaction. It eliminates the need for constant manual configuration by allowing users to add new capabilities via a simple plugin system.

How it works

Built on the Gemini Live API, the assistant uses native audio streaming for low-latency conversations and affective dialog, which allows it to sense emotion in the user's voice. It integrates a plugin architecture where users can drop .py files into a specific folder to add new skills without modifying the core engine. The system combines real-time screen capture and webcam vision with a suite of built-in actions for system control, web searching, and application management, all managed through a PyQt6-based HUD.

Who it’s for

Users looking for a JARVIS-style AI assistant that can control their OS (Windows, macOS, Linux) and integrate into their daily workflow through voice, vision, and extensible plugins.

Highlights

• Plugin System: Add new skills by dropping a single Python file into the plugins folder with built-in crash isolation.
• Affective Dialog: Adapts its tone based on the emotion detected in the user's voice.
• Proactive Audio: Distinguishes between direct commands and background chatter to avoid accidental triggers.
• System Control: Voice-driven control over hardware settings (volume, brightness), application launching, and desktop management.
• Visual Awareness: Real-time screen and webcam feeds piped into the Gemini session for visual context.
• Unlimited Sessions: Uses sliding-window context compression to maintain long conversations without losing the thread.