◵

Semantic Search

Search by meaning, not keywords. A real embedding model (all-MiniLM-L6-v2) runs in your browser — this is the retrieval step behind RAG.

</> Source

Loading the embedding model (~23 MB, first time only)…

Embeddings computed locally with 🤗 Transformers.js (all-MiniLM-L6-v2). Cosine similarity ranks the corpus by meaning. Nothing leaves your browser.