[DOC] Neural search does not support field vector_field

[savanbthakkar]

What do you want to do?

• Request a change to existing documentation
• Add new documentation
• Report a technical problem with the documentation
• Other

Tell us about your request. Provide a summary of the request and all versions that are affected.
As per Neural search documentation page here, neural query can have this field called vector_field. However, running a search with this field causes this error:
[neural] query does not support [vector_field]

What other resources are available? Provide links to related issues, POCs, steps for testing, etc.
I didn't find any similar issues or much documentation around neural search query.
I have 3 vector fields for each document that I index. Think of a product with 3 description fields. When I query with a keyword, I want to query all 3 vector fields.
This is how my query looks like:

GET /my_index/_search
{
  "_source": {
        "exclude": [ "doc_description1_vector","doc_description2_vector","doc_description3_vector" ]
    },
  "size": 50,
  "query": {
    "bool": {
      "should": [
        {
          "neural": {
            "doc_description1_vector": {
              "query_text": "Tennis Racquet",
              "vector_field": "doc_description1_vector",
              "model_id": "rTuZqYoB1bvGnxqnUifN",
              "k": 30
            }
          }
        },
        {
          "neural": {
            "doc_description2_vector": {
              "query_text": "Tennis Racquet",
              "vector_field": "doc_description2_vector",
              "model_id": "rTuZqYoB1bvGnxqnUifN",
              "k": 30
            }
          }
        },
        {
          "neural": {
            "doc_description3_vector": {
              "query_text": "Tennis Racquet",
              "vector_field": "doc_description3_vector",
              "model_id": "rTuZqYoB1bvGnxqnUifN",
              "k": 30
            }
          }
        }
      ]
    }
  }
}

And here is the response in OpenSearch Dashboard dev tools.

{
  "error": {
    "root_cause": [
      {
        "type": "parsing_exception",
        "reason": "[neural] query does not support [vector_field]",
        "line": 13,
        "col": 31
      }
    ],
    "type": "x_content_parse_exception",
    "reason": "[13:31] [bool] failed to parse field [should]",
    "caused_by": {
      "type": "parsing_exception",
      "reason": "[neural] query does not support [vector_field]",
      "line": 13,
      "col": 31
    }
  },
  "status": 400
}

[hdhalter]

@macohen - Is there someone that might be able to test this?

[macohen]

maybe @navneet1v or @vamshin?

[navneet1v]

Hi @savanbthakkar,

You don't need to specify the vector_field as a parameter. Here is the updated query below.

GET /my_index/_search
{
  "_source": {
        "exclude": [ "doc_description1_vector","doc_description2_vector","doc_description3_vector" ]
    },
  "size": 50,
  "query": {
    "bool": {
      "should": [
        {
          "neural": {
            "doc_description1_vector": {
              "query_text": "Tennis Racquet",
              "model_id": "rTuZqYoB1bvGnxqnUifN",
              "k": 30
            }
          }
        },
        {
          "neural": {
            "doc_description2_vector": {
              "query_text": "Tennis Racquet",
              "model_id": "rTuZqYoB1bvGnxqnUifN",
              "k": 30
            }
          }
        },
        {
          "neural": {
            "doc_description3_vector": {
              "query_text": "Tennis Racquet",
              "model_id": "rTuZqYoB1bvGnxqnUifN",
              "k": 30
            }
          }
        }
      ]
    }
  }
}

The general syntax of the query clauses in Opensearch is:

<QUERY_CLAUSE_NAME> : {
   <FIELD_NAME>: {
      <DIFFERENT_REQUIRED_PARAMETERS_KEY>: <VALUE>
   }
}

I hope this clarifies and resolves your issue. If this is not what you are seeing in documentation please let us know, we will fix the documentation.

[hdhalter]

Hi @kolchfa-aws , tagging you for visibility, in case there is anything we can clarify in the documentation. Thanks.