How to retrieve all possible values of an attribute in metadata

[OweysMomenzada]

Is there a way to retrieve all possible values of a metadata? Or is that not possible?
Lets say I have the following metadata structure for all entries inside Qdrant (with three example entries):

{
 'page_content': <some text>.
  'metadata': {
      "date-of-upload": "2023-09-24",
      "language": "english",
      "page": 15,
      "source": "english/sensors-21-01322-v2.pdf"
    }
}


{
 'page_content': <some text>.
  'metadata': {
      "date-of-upload": "2023-07-14",
      "language": "english",
      "page": 11,
      "source": "english/sensors-21-01322-v2.pdf"
    }
}


{
 'page_content': <some text>.
  'metadata': {
      "date-of-upload": "2023-01-24",
      "language": "english",
      "page": 20,
      "source": "english/earning_calls.pdf"
    }
}

Now I want to now if there is a possibility to query all unique values for source:
>>> ["english/earning_calls.pdf", "english/sensors-21-01322-v2.pdf"]

[timvisee]

No. Qdrant does not support such feature to get unique values.

You could scroll over all points, extract the desired metadata, and only keep unique values on your client. This would be very inefficient though.

[djptek]

Hi @OweysMomenzada

I'd just like to update you on the availability of the facet api which might fulfil your requirement

Here is a brief example of how that works

PUT collections/test
{
  "vectors": {
    "size": 3,
    "distance": "Dot"
  }
}

PUT /collections/test/index
{
    "field_name": "name",
    "field_schema": "keyword"
}


PUT collections/test/points
{
  "points": [
    {
      "id": 1,
      "vector": [255,0,0],
      "payload": {
        "name": "RED"
      }
    },
    {
      "id": 2,
      "vector": [0,255,0],
      "payload": {
        "name": "GREEN"
      }
    },
    {
      "id": 3,
      "vector": [0,0,255],
      "payload": {
        "name": "BLUE"
      }
    }
  ]
}

POST /collections/test/facet
{
    "key": "name"
}

### response

{
  "result": {
    "hits": [
      {
        "value": "BLUE",
        "count": 1
      },
      {
        "value": "GREEN",
        "count": 1
      },
      {
        "value": "RED",
        "count": 1
      }
    ]
  },
  "status": "ok",
  "time": 0.000356093
}

[FilippTrigub]

+1

having the same issue right now

[timvisee]

Please consider using the Facet API introduced in Qdrant 1.12.

Ref: https://github.com/orgs/qdrant/discussions/2721#discussioncomment-11567994

Docs: https://qdrant.tech/documentation/concepts/payload/#facet-counts