Ternlight – 7 MB embedding model that runs in browser (WASM)
Summary
Ternlight presents a 7 MB embedding model that runs entirely in the browser (WASM) with CPU-only performance and no API calls. It ships as a single npm package, emphasizing self-contained, serverless embeddings and in-browser semantic search. The article demonstrates use cases like a React docs search and notes MIT licensing, highlighting a privacy-preserving, lightweight alternative to cloud-hosted embeddings.