Embeddings
@kortexya/reasoninglayer / Embeddings
Embeddings
Section titled “Embeddings”Embedding-space ranking — POST /api/v1/embeddings/rank.
Scores candidate texts against a query by cosine similarity in the deployment’s shared sentence-embedding space. The endpoint never reorders and never truncates: it reports one similarity per candidate, in request order, and leaves the ranking policy to the caller.
Distinct from the RAG search in types/rag.ts (which retrieves from the
knowledge base) — nothing here reads or writes the KB.
References
Section titled “References”EmbeddingRankRequest
Section titled “EmbeddingRankRequest”Re-exports EmbeddingRankRequest
EmbeddingRankResponse
Section titled “EmbeddingRankResponse”Re-exports EmbeddingRankResponse