Shard

A Tauri application for easy interaction with AI models.

Note: Control (^) + Space is the shortcut to toggle visibility of Shard, and Control + K is the shortcut for OCR capture.

Features

• Chat Interface with Streamed Responses
• Free Models (DeepSeek V3, Gemini Flash, DeepSeek R1)
• API Key Management (OpenRouter & Google)
• Screen Capture + OCR Analysis (Built-in) + Image Copy Paste
• Markdown & LaTeX Rendering
• Background Panel Mode (MacOS only)
• System Prompt
• Reasoning Data for R1 & Reasoning Summaries for Gemini
• MCP Tool Reasoning Integration - AI models make smarter tool choices
• Stock Price Data from Yahoo Finance
• Weather Data from Open-Meteo
• General Data from Wikipedia
• Wikipedia Research Loop (up to 4 searches)
• Research Paper Data from ArXiv

MCP Tool Reasoning

Shard includes an advanced Model Context Protocol (MCP) integration that enhances AI decision-making for tool usage. The system provides:

• Intelligent Tool Selection: AI models receive structured guidance on when and how to use research tools
• Context-Aware Hints: Real-time suggestions based on query analysis (weather, financial, academic research, etc.)
• Multi-Tool Research Strategies: Guidance for combining tools effectively (e.g., Wikipedia for background + ArXiv for latest research)
• Transparent Reasoning: AI explains its tool selection process to users

See MCP_INTEGRATION.md for detailed documentation.

Feel free to add more ideas in the Issues or contribute to the project with a PR!

Prerequisites

Shard comes with built-in OCR capabilities and doesn't require any additional system installations. The application includes all necessary components for text recognition.

Development

To build and run the application in development mode:

1. Clone the repository:

   git clone <repository_url> # Replace with actual URL after publishing
   cd shard

2. Install Rust and Node.js/npm if you haven't already. Refer to the official Rust and Node.js websites.
3. Install Tauri prerequisites: Follow the guide on the Tauri website.
4. Install project dependencies:

   cargo fetch # Fetches Rust dependencies
   # Frontend dependencies might be needed depending on the chosen frontend setup
   # e.g., npm install if using a Node.js-based frontend

5. Run the development server:

   cargo tauri dev

To build the application for production:

cargo tauri build

This will generate installers/packages in the src-tauri/target/release/bundle/ directory.

License

This project is licensed under the MIT License.