Vector Search schema problem!

MongoDB Atlas Vector Search
1

Hi!
I followed this tutorial and it worked really well for me:

The problem I have is that the text I want to store is too big to be represented in a single vector. Now I want to store the original (full) text along with several vectors in one document.
Then, when I use vector search and one of the vectors matches I want to get a match in my query.

Is that possible? If yes, how?

3

Hi @Patrick_Treppmann and welcome to MongoDB community forums!!

If I may ask, how big is the text exactly and are you being returned with any specific error messages whilst trying to create the vectors for this text?

As per my understanding from the above statement, you are trying to create two vector embedding fields and create vector index on both of them. Though vector search allows you to create index on multiple fields like:

{
  "mappings": {
    "dynamic": true,
    "fields": {
      "vector1": {
        "dimensions": 384,
        "similarity": "euclidean",
        "type": "knnVector"
      },
      "vector2": {
        "dimensions": 384,
        "similarity": "euclidean",
        "type": "knnVector"
      }
    }
  }
}

considering vector1 and vector2 are two vector embedded fields, but the catch here is, only one query operator is allowed under $search - and knnBeta specifically is only allowed at the top-level, so it can’t be nested in a compound either. So, one at a time.

I hope I was able to help you with your query.

Warm regards
Aasawari