Vector Atlas search Embedded documents

i have a collection with documents structured this way

{
  "_id": {
    "$oid": "65a84ffc0000000000000000"
  },
  "lyrics": [
    {
      "line":  "hello",
      "embedding": [ ]
    }
  ]
}

I want to search by line, I created the following Atlas Search index for this collection.

{
  "mappings": {
    "dynamic": false,
    "fields": {
      "lyrics": {
        "dynamic": false,
        "fields": {
          "embedding": {
            "dimensions": 1536,
            "similarity": "euclidean",
            "type": "knnVector"
          }
        },
        "type": "document"
      }
    }
  }
}

the status of this index been like this for a while, am I doing something wrong here

Hi Kata,

Thank you for submitting your question. Unfortunately we don’t presently support searching against nested vectors. We have this on our roadmap, and I will make sure to update this thread when it’s supported.

If there is any metadata in the parent document that you want to filter on, you can always use $lookup and $match to join against the parent document using a field in the child document, and then post-filtering on any metadata present within the parent. Happy to provide an example of this if that’s what you are looking for.

Hi Henry,
Just wondering if you might have an estimate for when this feature (searching against nested vectors) will become available?
Thanks Allen

1 Like

HI @Allen_Lee,

Hopefully towards the end of this year we will support some form of query patterns against a nested data model, though the design discussions for this are still in flight. I will update on this thread when we have a clearer idea of when this will land, as we know it is something we need to provide.

1 Like