Searchable Screenshots

A Python application that indexes and enables fast searching of your screenshots using OCR, vision models, and vector embeddings.

⚠️ Under active development. Use at your own risk.

Features

• Folder scanning – Recursively scan selected folders for image files.
• OCR extraction – Pull text from screenshots via Tesseract.
• Vision description – Generate concise image captions using a local LLM (e.g., Gemma3).
• Embeddings – Create vector embeddings for hybrid text‑image search.
• Hybrid search – Combine keyword and visual similarity for powerful queries.
• Configurable parallelism – Adjust concurrency for indexing performance.

Installation

This project uses uv package manager, install it at https://docs.astral.sh/uv

# Clone the repository
git clone https://github.com/Creative-Geek/Searchable-Screenshots.git
cd Searchable-Screenshots

# Install dependencies and create virtual environment
uv sync

Usage

Run the GUI

uv run main.py

The application will open a window where you can add folders, configure settings, and start indexing.

Index from the command line

uv run src.core.processor --folder "C:\path\to\screenshots"

This will process the images and populate the local database.

Configuration

Open Settings in the GUI to adjust:

• Ollama URL
• Vision model name
• Embedding model name
• Parallel processing count (default = 1)

Contributing

Feel free to open issues or submit pull requests. Please follow the existing code style and run the test suite before submitting.

License

This project is licensed under the MIT License.