Charon: Autonomous Desktop Assistant

Charon is an autonomous desktop assistant designed to simplify user actions through local execution and artificial intelligence. Built with PyQt6 and powered by the Gemini Live API, it provides real-time voice and text interaction, screen perception, and desktop control entirely from your computer.

Core Capabilities

• Low-Latency Voice and Text: Interactive speech-to-speech interaction using Gemini 2.5 Live models.
• PyQt6 Interface: Interactive dashboard featuring real-time CPU, Memory, GPU, and Network metric monitoring.
• Vision and Screen Processing: Local screenshot and webcam frame capture analyzed directly through Gemini's vision capability.
• Desktop Automation: Simulates keystrokes, mouse operations, window focuses, and file operations.
• Browser Automation: Built-in Playwright integration to open web browsers, search, click, type, and extract data.
• Task and Code Agent: Specialized modules to write, edit, run, and build code files or handle multi-step actions.
• Persistent Memory: Local memory system tracking preferences, active projects, and user notes.
• Game Hub Integration: Interacts directly with local Steam and Epic Games clients to manage updates.

Architecture

A comparison of performance metrics across different scenarios:

Setup and Installation

Prerequisites

• Python 3.11 or 3.12
• Windows Operating System
• Gemini API Key

Installation Steps

1. Create a virtual environment:

   python -m venv venv

2. Activate the virtual environment:

   venv\Scripts\activate

3. Run the installer script to download dependencies and configure Playwright:

   python setup.py

Configuration

1. Copy the template API key file: Copy config/api_keys.json.template to config/api_keys.json

2. Edit config/api_keys.json and replace YOUR_GEMINI_API_KEY_HERE with your actual Gemini API key:

   {
       "gemini_api_key": "YOUR_GEMINI_API_KEY_HERE",
       "os_system": "windows"
   }

Execution

Ensure your virtual environment is active, then start the assistant:

python main.py