Retrieve the right answer, not just semantically similar ones. Encode temporal, topical, and other dimensions into existing embeddings with one API call.