Show HN: Photon – Rust pipeline that embeds/tags/hashes images locally w SigLIP (github.com)

Photon, an AI-powered image processing pipeline written in Rust, has been announced, enabling local image analysis without reliance on cloud services. Utilizing the SigLIP model, Photon processes images to produce structured JSON outputs that include 768-dimensional vector embeddings, semantic tags, EXIF metadata, content hashes, and thumbnail images. This design allows users to operate entirely offline, making it suitable for environments where data privacy and control are paramount. The significance of Photon lies in its pure processing pipeline that eliminates the need for databases or external servers, thus simplifying integration into local storage or search infrastructures. Key features include zero-shot tagging with a vocabulary surpassing 68,000 terms, high-performance embedding via ONNX Runtime, and batch processing capabilities with progress tracking. The single binary architecture, supporting macOS and Linux, offers developers an accessible and efficient way to embed advanced image processing into their Rust applications, bolstering the toolkit available for AI/ML developers focused on local data handling and security.