Client-side semantic search for your static site
Summary
The article shows how to do client-side semantic search for static sites, weighing a 4 MB lookup-table model against larger in-browser transformers. It covers building an index via chunking, quantization to int8, a JS WordPiece tokenizer, and a pragmatic approach to ranking with Reciprocal Rank Fusion. A practical Python toolchain and GitHub code are provided for readers to experiment locally.