Hachi: An Image Search Engine
12 days ago
- #self-hosted
- #machine-learning
- #search-engine
- Hachi is a self-hosted search engine designed to search personal data across various devices and storage locations.
- Current implementation focuses on image search, with future plans to support video, text, and audio.
- The project emphasizes minimal external dependencies, hackability, and experimentation with semantic search.
- Features include face recognition, meta-data indexing, and vector search for semantic queries.
- Backend is written in Python and Nim, with a frontend built using HTML, JS, and CSS.
- The system supports distributed queries and aims for high performance on diverse hardware.
- Machine learning models are integrated for tasks like face clustering and semantic embedding.
- The project is tested on Windows and Linux, with potential for extension to other platforms.
- Philosophy includes learning from open-source, optimizing for personal use, and resisting data duplication.
- Future goals include supporting remote storage indexing and improving user interfaces.