createEmbeddingRetriever
createEmbeddingRetriever(
documents,embed,options?):Promise<Retriever>
Defined in: retrievers/embedding-retriever.ts:64
Build an in-memory vector retriever that scores documents by cosine similarity against an embedded query. Documents are embedded once up front, so repeated queries reuse the cached vectors.
This is a deterministic, dependency-free way to evaluate embedding quality:
inject a real provider for live scoring, or a fixed stub for reproducible
tests. The same embed function must be used for documents and queries.
Parameters
Section titled “Parameters”documents
Section titled “documents”readonly EvalDocument[]
options?
Section titled “options?”EmbeddingRetrieverOptions = {}
Returns
Section titled “Returns”Promise<Retriever>