What is semantic search?
Semantic search embeds a natural-language query and returns the nearest stored chunks by vector similarity, each with its similarity score and its source document and URL. It is a lookup, not a generation: no model writes an answer, so the call is flat-priced.
Why it matters
Keyword search fails exactly where a user’s wording differs from the document’s. Returning provenance with every chunk is the other half of what makes it usable — a retrieval result you cannot cite is a retrieval result you cannot ship.