Should a RAG system change databases just to add hybrid search?

A working RAG system is being upgraded with . The current setup uses Chroma as the , MMR retrieval, , , LangChain, and Ragas for evaluation. The existing evaluation already checks context precision, context recall, , and answer relevancy.

There are three possible paths. LangChain EnsembleRetriever can keep Chroma in place and add BM25 plus dense search with RRF fusion in only a small code change. Moving to Weaviate or Qdrant would give native in one call, with an alpha setting to tune the mix, but it would require leaving Chroma and re-ingesting all data.

LlamaIndex QueryFusionRetriever is another clean fusion option, but it means moving away from LangChain. The real question is whether EnsembleRetriever is good enough in , whether native brings a clear quality or latency gain, and whether adding a reranker may improve retrieval more than itself.

Key points

  • The current RAG stack uses Chroma, LangChain, , , and Ragas evaluation.
  • The smallest change is LangChain EnsembleRetriever, which mixes BM25 and dense search while keeping Chroma.
  • Weaviate and Qdrant offer native , but require migration and re-ingestion.
  • LlamaIndex QueryFusionRetriever is another option, but it changes the framework choice.
  • The useful test is whether retrieval scores and latency improve enough to justify more engineering work.
Read original