Show HN: ParqDB – Vector search in the browser from Parquet over HTTP"

ParqDB enables vector search directly in the browser by hosting Parquet-indexed data on object storage, allowing users to search 100,000 Wikipedia articles without sending queries to a server. The system uses IVF-LVQ8 indexing and WebAssembly to embed queries locally and retrieve only necessary byte ranges via HTTP, ensuring query privacy while eliminating the need for a dedicated vector database server. This approach combines browser-side embedding and ranking with efficient object storage access patterns for fast, privacy-preserving semantic search.