Request/Question Vector Indexes with Microsoft Kernel-Memory and Semantic-Kernel

[JohnGalt1717]

I see the documentation that's available for the vectorization index that are being created for more like this.

This brings up an interesting solution to a problem, in that we have content in RavenDb and I'd love to be able to use Kernel-Memory against our AIs with RavenDb to store the vectors and have RavenDb be able to do the lookups that Kernel-Memory and Semantic-Kernel support. (i.e. AI Document search and response using RAG)

Is there a way that I can pass in the embeddings for these indexes and just execute a command to generate these ad-hoc?

Specifically, I'm looking at implementing RavenDB as an IVectorDb like Qdrant does. I think it would be great if there was an integration into all of this stuff for RavenDb and would be a great additional selling point to be able to use AI Enrichment to create the vectors and then search on them etc.

I'd be happy to create this with a little guidance, but from what I can tell I can't see a way to be able to actually pass in the vectors for each document to the indexes.

For reference here's the Qdrant version of IVectorDb: https://github.com/microsoft/kernel-memory/blob/a26407b972a7e61d86f3657b6ac6d8281ffffcab/dotnet/CoreLib/MemoryStorage/Qdrant/QdrantMemory.cs#L15

Thanks!

[maciejaszyk]

Hi,
a very similar question was asked here: #16603

[JohnGalt1717]

Ok. Well, I guess what I'm asking is for the next release of RavenDb to be VectorDb compliant like Cosmos and the others so we don't have to also run a separate vector database and both do its own vectoring and allow embeddings from other LLM models.

This would be killer right now.