rclip is a command-line tool that leverages OpenAI's CLIP model for AI-powered semantic search and similarity matching of images. It allows users to find images based on text descriptions or by comparing them to other images, making it useful for researchers, developers, and power users managing large image collections.

How It Works

rclip utilizes CLIP's multimodal embedding capabilities to represent both images and text in a shared vector space. When searching, it converts the query (text or image) into a vector and then finds images whose vectors are closest in this space. This approach enables nuanced, context-aware image retrieval beyond simple keyword matching. The tool can also combine multiple text and image queries using boolean logic for more refined searches.

Quick Start & Requirements

• Linux: sudo snap install rclip or download AppImage from releases.
• macOS: brew install yurijmikhalevich/tap/rclip
• Windows: Download MSI from releases.
• Pip (all OS): pip install --extra-index-url https://download.pytorch.org/whl/cpu rclip (CPU-only PyTorch is used by default).
• Prerequisites: Python, Poetry (for local development). Initial indexing of images can be time-consuming, depending on CPU and dataset size.
• Docs: Blog, Demo

Highlighted Details

• Supports text-to-image and image-to-image search.
• Allows combining queries with boolean logic (AND, OR, NOT).
• Offers terminal preview capabilities for iTerm2, Konsole, wezterm, Mintty, and mlterm via --preview.
• Can output file paths only for integration with external viewers.

Maintenance & Community

• Follows Conventional Commits standard.
• Contributions are welcome.
• Discussions can be initiated via GitHub issues.

Licensing & Compatibility

• MIT License. Permissive for commercial use and integration with closed-source projects.

Limitations & Caveats

• The default pip installation uses a CPU-only PyTorch build, which may be significantly slower for feature extraction and search compared to a GPU-accelerated setup.
• Terminal preview functionality is limited to specific terminal emulators.