similarity-search-kit

Why we included this project

If you build iOS or macOS apps that need to search over local documents, this Swift package gives you a working semantic search layer without sending text to a remote API. It wraps several sentence-embedding models, including Apple's built-in NaturalLanguage embeddings and quantized MiniLM/DistilBERT variants, behind a small SimilarityIndex API, so you can index text, query it, and get ranked results with a few lines of code. The on-device design matters for privacy-sensitive use cases like searching confidential files or building offline question-answering over a local dataset, and the protocol-based architecture lets you swap in your own embedding model or distance metric. Teams already invested in Swift and CoreML will find it a natural fit, while the included example apps (PDF search, chat-with-files) show realistic integration patterns rather than toy snippets.