I added semantic search to my React app without a backend (and it's under 1ms)

Short summary

Built altor-vec, a 54KB WASM-compiled vector search library that runs entirely in the browser with zero per-query cost, achieving <1ms p95 query latency for 10K vectors. Uses free HuggingFace embeddings and HNSW indexing to deliver semantic search (understanding meaning, not just text matching) on static indexes. Ideal for documentation sites and marketing sites where content is public and updates at deploy time; tradeoffs exist for real-time content, millions of documents, and private data.

• •54KB WASM library runs semantic search entirely in browser with <1ms latency
• •Free embeddings + HNSW indexing; zero per-query cost vs. Algolia
• •Best for static content (docs, marketing); not suitable for real-time or private data

Generated with AI, which can make mistakes.