The community member is asking if it is possible to use a Postgres vector store (pgvector) as a docstore to build a QueryFusionRetriever with a vector_retriever and a bm25retriever. They note that the data is already stored in the pgvector, but the bm25retriever can only be initialized with a docstore. The community member also mentions finding a sparse vector_store_query_mode, but believes it is only for hybrid text search, which is not what they are looking for. The comments suggest that the bm25retriever can be initialized with a list of nodes, and there is a question about how to get all the nodes from a vector store and whether it is possible to retrieve with a top k of 10000.
Hi, is there a way to use a postgres vector store (pgvector) as a docstore? All the data is actually already there.
Specifically, what I want to do, is use the data we already store in pgvector to build a QueryFusionRetriever, with a vector_retriever and a bm25retriever. In theory, I should be able to perform bm25 retrieval in the postgres database, but the bm25retriever can only be initilised with a docstore.
I also found out this sparse vector_store_query_mode, but I believe thats only for hybrid text search, so not what I am looking for.
