Developer builds sub-millisecond semantic search for React apps with no backend
A developer has built a browser-native semantic search tool called altor-vec that runs entirely client-side using WebAssembly, eliminating the need for a server, API keys, or per-query costs. The tool uses HNSW vector search compiled to a 54KB WASM module, paired with HuggingFace embeddings generated at build time rather than at query time. Unlike fuzzy-text libraries such as Fuse.js, the solution understands meaning, allowing a query like 'cancel my plan' to match a document titled 'end your subscription.' The search index is built once during deployment as a static binary file served from a CDN and loaded into the browser on first use. The developer shared a full React code walkthrough, noting query response times of under one millisecond after the initial model load.
This is an AI-generated summary. ShortSingh links to the original source for the complete article.
Discussion (0)
Log in to join the discussion and vote.
Log in