openviking

Hybrid search combines vector similarity matching with BM25 keyword scoring to improve document retrieval quality. You need it when keyword matching alone fails to find semantically related content in large knowledge bases.

To retrieve passages using semantic search, request a search for relevant content about a specific topic. The system applies vector similarity matching and BM25 scoring to locate and rank the most pertinent information from your stored documents.

Yes, OpenAI embeddings access is required to perform vector similarity matching. The system relies on these embeddings to evaluate semantic similarity and retrieve related content based on meaning rather than exact terms.

Combining BM25 and vector search for RAG workflows is best achieved through hybrid retrieval. This approach merges keyword scoring with semantic matching, enabling filtered retrieval and result reranking to find the most relevant passages.

Yes, you can apply filtering and reranking to memory store search results. The system supports these features alongside vector similarity matching and BM25 scoring to refine outputs for retrieval-augmented generation workflows.

Semantic search returns better results because it retrieves related content based on meaning rather than exact terms. By integrating BM25 keyword scoring with vector search, hybrid retrieval captures both semantic similarity and precise keyword matches.