Possibility of Replacing parquet files with Graph database like Neo4j during indexing

[amitt0488]

Is there a way by which we can make sure indexing stores data in a GraphDB like Neo4j instead of generating parquet files? Doesn't look scalable that parquet files keep on generating in the root folder.
Is this currently supported or not?

[ntsarb]

According to the docs, the most recent versions of GraphRAG use Azure AI Search and LanceDB (default) as vector store. I can see the cosmos container, azure.storage.blob and the azure.search.documents.indexes in the source code. I suppose it will take a significant effort to develop a common API and adapter classes for compatibility with different vendors/components, but it's worth the effort; the more options, the wider the applications and the adoption.