rag-hybrid-search-definition

Summary

The system's search pipeline doesn't rely on a single lookup method; it pulls results from both keyword matching and meaning-based (vector) search, then blends and re-ranks them so the LLM receives a tighter, more relevant set of context. This matters because it means answer quality depends on both retrieval paths working together, and any gap in one can be compensated by the other before the model even sees the input.