I am making an AI assistant (RAG) that answer questions from mongodb atlas database.
the database has many collections but the problem is that i can only create vector search index on one collection but i want to make one for all the database
i searched for a solution and i saw that the only solution is to combine all the database into one collection but i don’t want to replicate the database
I am also facing same issue. Is their any other option ? please reply on this.
Hi @ADEEB_Alkahat! Thank you for the question.
We are planning on adding support for building search and vector search indexes against views of collections, which may eventually include views built from multiple collections via $lookup.
In the meantime, can you provide more details about the data model? It might be that the best approach is to ensure multiple vector search indexes are built, each against their own collection, with metadata separation occurring via the index naming scheme as shown below:
collection1: vs_index_1,
collection2: v2_index2,
...
Provided you have less than a few hundred indexes you are trying to build, the load on the database produced by the changestream opened by each search index should be manageable.