Question: Can we specify custom vector databases at kg index creation?

[rasharab]

You've provided custom vector retrievers, but trying to ascertain if it is currently possibly to use pinecone (for example) during KG index creation (i.e. embedding relationships etc.)

I'm assuming you guys are storing entity relationship embeddings etc. in your library using neo4j vectorstore?

Also, there's a lot of functions under indexes.py (upsert_vector_on_relationship ...) that I'm not entirely sure if they're even being used in this package.

Thank you again.

[stellasia]

Hi @rasharab ,

You're right in assuming that the current implementation of the KG creation pipeline writes all results (nodes, relationships and embeddings) to Neo4j. If you want to change this behavior, you can create a custom KGWriter.

Regarding the functions in indexes.py, they are helper functions not used directly by other components of the package, but that users can call if they can't write Cypher queries.

Hope that helps!