Merging BM25 and Vector Retrieval in Llama-Index Workflow
Merging BM25 and Vector Retrieval in Llama-Index Workflow
At a glance
The community member is asking if it is possible to implement the capabilities of KnowledgeGraph (Grafoù Properties) in a workflow that already uses QueryFusionRetriever, by merging BM25 and vector retrieval. Another community member responds that they can combine any retrievers in the query fusion retriever. The original community member then clarifies that they are building a RAG (Retrieval-Augmented Generation) application that uses a hybrid of BM25 and vector retrievers, stores data in Chroma DB, and uses Cohere's reranking as a node post-processor. They mention that their current setup lacks context, and they would like to implement KnowledgeGraph to address this. The community member then asks if it is possible to implement three retrievers, with Chroma as the database, in the QueryFusionRetriever and pass it to Cohere's reranker.
hello everyone ! for llama-index, is it possible to implement the capabilities of KnowledgeGraph (Grafoù Properties) in a workflow that already uses QueryFusionRetriever, by merging bm25 + vector retriving ?
Sorry, i misspoke. simply, i am building a rag-app that uses the hybrid BM25 + Vector as retrievers, store with Chroma db and use Cohere's reranking as nodepostprocessor. i use textbook chapters (biology, pathology) as knowledge for RAG. it is obvious that it lacks context when responding. i would like to implement KnowledgeGraph to fill this lack of context. so the question is: is it possible to implement three retrievers, with Chroma as db, in the QueryFusionRetriever and pass it to Cohere's reranker ?
