Snapy

AI-powered screenshot manager with intelligent classification, search, and analytics.

Features

• Smart Classification: CLIP (ViT-B/32) automatically categorizes screenshots into Person, Documents, Code, Browser, Chat, Games, and Other
• OCR Search: Extract and search text from screenshots (English + Arabic)
• Analytics: Detailed statistics about your screenshot collection
• Privacy-First: Everything runs locally, completely offline after setup
• Fast: ONNX runtime for efficient CPU inference

Quick Start

Download Release (Recommended)

# Linux
wget https://github.com/Tojan-Naiem/snapy/releases/latest/download/snapy-v1.1.0-linux-x64.zip
unzip snapy-v1.1.0-linux-x64.zip
cd linux-x64
./setup.sh

# Windows
# Download snapy-v1.1.0-win-x64.zip from releases
# Extract, install Tesseract OCR, run setup.ps1

Build from Source

git clone https://github.com/Tojan-Naiem/snapy.git
cd snapy
./setup.sh

Usage

snapy organize ~/Screenshots    # Categorize screenshots
snapy search "invoice"           # Search by text content
snapy stats ~/Screenshots        # View statistics
snapy info screenshot.png        # File metadata
snapy restart ~/Screenshots      # Undo categorization

How It Works

• Classification: CLIP model converts images to embeddings and matches against category embeddings
• Search: Tesseract OCR extracts text, stores in SQLite for fast full-text search
• Performance: ~100-200ms per classification, near-instant search

Requirements

• Linux: Ubuntu/Debian, Python 3.8+
• Windows: Windows 10/11, Python 3.8+, Tesseract OCR
• .NET 8.0 SDK (if building from source)

Technical Stack

C# (.NET 8.0) • CLIP (ONNX) • Tesseract OCR • SQLite • ImageSharp

Architecture

Snapy.Cli/              # CLI interface
Snapy.Core/             # Domain entities
Snapy.Infrastructure/   # AI models, database, services
Models/                 # ONNX models and embeddings

Known Limitations

• Categories fixed at compile-time
• No progress bars for batch operations
• Primary testing on Ubuntu/Debian

Contributing

Contributions welcome! Submit issues or pull requests.

License

MIT License - see LICENSE file for details.

Acknowledgments

Built with CLIP, Tesseract OCR, and ONNX Runtime

---

Privacy Notice: All processing happens locally. No data sent to external servers.