FusionRetrieverOptions
Properties
Section titled “Properties”bm25: BM25Index;The BM25 index for keyword search.
bm25Weight?
Section titled “bm25Weight?”optional bm25Weight?: number;Weight for BM25 scores in the fused result. The vector weight is
1 - bm25Weight. Default: 0.3 (70% vector, 30% keyword).
embedder
Section titled “embedder”embedder: Embedder;The embedder to convert query text to vectors.
fetchK?
Section titled “fetchK?”optional fetchK?: number;Number of candidates to fetch from each source before fusion. Should be >= topK. Default: topK * 3.
indexName
Section titled “indexName”indexName: string;The index name in the vector store.
reranker?
Section titled “reranker?”optional reranker?: Reranker;Optional post-retrieval reranker.
optional topK?: number;Default number of results to return. Default: 10.
vectorStore
Section titled “vectorStore”vectorStore: VectorStoreCore;The vector store for semantic search.