Sentence embeddings in the browser
A quantized MiniLM sentence-embedding model runs in your browser and
scores how similar your sentences are in meaning — not spelling. Each
line below becomes a 384-dimensional vector; pairs are ranked by
cosine similarity. Your text is embedded on your own machine and never
sent anywhere.
The cat sat on the mat.
A feline rested on the rug.
The stock market fell sharply today.
Equities dropped steeply this afternoon.
I love programming in Python.
Runtime: Transformers.js + ONNX Runtime (WebGPU with WASM fallback),
model all-MiniLM-L6-v2 (q8), all pinned and served from
/vendor/ and /models/. See the
colophon.