AI / LLM

Vector / Embedding Lab

Paste two embedding vectors — comma, space, newline or JSON-array separated — and read the metrics that drive semantic search and RAG retrieval. The angle gauge shows how aligned the two vectors are.

Vector Adim 0

Vector Bdim 0

Cosine similarity−1 opposite · 0 orthogonal · 1 identical

—

Angle betweenarccos of cosine

—

Dot productΣ aᵢbᵢ

—

Euclidean distance‖a − b‖₂

—

cosine similarity

−10+1

‖a‖ = — · ‖b‖ = —. Cosine similarity is the workhorse of vector search because it ignores magnitude and compares direction only — ideal when embeddings aren't normalized. Everything is computed locally; vectors never leave this page.