Qdrantvectorstore: Support for Multiple Dense Vectors in Qdrant

Product Help
1

Hi everyone,

I’ve been working with QdrantVectorStore in LangChain and ran into a limitation: the current integration only supports one (default) dense vector field (plus optionally one sparse vector for hybrid).

However, Qdrant collections can be configured with multiple dense vector fields. Right now, LangChain’s QdrantVectorStore can’t handle this setup — vector_name is a single string, so you can’t index or search across multiple dense fields.

What I’d like to propose

  • Replace self.vector_name with self.vector_names (a list).

  • When adding documents, generate embeddings for each field in the list.

  • For retrieval allow queries across all dense vector fields. Get result for each field and then merge results using fusion strategy (e.g., RRF). For Hybrid Search we would there would be 2 steps of fusion.

This would make the integration consistent with Qdrant’s actual capabilities and unlock more powerful retrieval strategies directly from LangChain.

Would love to hear your thoughts — I already have a draft implementation, would it be a good idea if I put together a PR with these changes ?