Vector database use for RAG

[MrCsabaToth]

I'm wondering if there are any examples for RAG scenarios. How would that fit into the Koog workflow, and what databases people use?

[davidb-e4s]

To answer your second point...we're using MongoDB to store our vectors. Is it the 'best' vector database? Probably not in raw terms but the value of having one data platform is worth it. Plus Mongo just acquired VoyageAI so they seem to be committed to driving this forward. They also now support scalar and binary indexing quant out of the box which is a huge plus in my book.

Disclaimer: I am a MongoDB fanboy

[MrCsabaToth]

MongoDB is certainly a great database, however I am looking into on-device edge RAG, specifically on a very small Android device. That's where Objectbox comes into picture.

[davidb-e4s]

Sounds like a pretty neat use case! I hope you get it working :)

[Ololoshechkin]

Hi! We have https://docs.koog.ai/ranked-document-storage/ in Koog but specific support for vector databases is still in progress, in the meanwhile -- feel free to wrap any vector database client into that Koog interface.

We're also working on integration with Spring AI that particularly has support for 20+ vector databases (so no need to re-implement the wheel, only provide a wrapper).

[noahkohrs]

Hello!

When can we expect the wrappers to be added ?

Thank you in advance 😄