[BUG] Neural query modelId is not nullable when used in nested query

[marcus-bcl]

Describe the bug

Using the neural_query_enricher processor in a search pipeline allows you to set the default model_id for a neural query.

However, when trying to use a nested query containing a neural query (e.g. for chunked text embeddings), it still requires you to provide the model_id.

Related component

Search

To Reproduce

1. Create an indexed with a nested knn field.
2. Add a search pipeline with a default_model_id, as per https://opensearch.org/docs/latest/search-plugins/semantic-search/#setting-a-default-model-on-an-index-or-field
3. Perform a nested query e.g.

GET /my-index/_search
{
  "query": {
    "nested": {
      "score_mode": "max",
      "path": "text_embedding",
      "query": {
        "neural": {
          "text_embedding.knn": {
            "query_text": "my query text"
          }
        }
      }
    }
  }
}

4. 500 Error is returned:

[HTTP/1.1 500 Internal Server Error]
{"error":{"root_cause":[{"type":"null_pointer_exception","reason":"modelId is marked non-null but is null"}],"type":"null_pointer_exception","reason":"modelId is marked non-null but is null"},"status":500}

Expected behavior

Expect a 200 response with search results

Additional Details

Plugins
ml

Host/Environment (please complete the following information):

• Amazon OpenSearch 2.13

[getsaurabh02]

@vibrantvarun do you want to verify this and take a stab?

[msfroh]

I think this was fixed by #14739 (opened in response to opensearch-project/neural-search#813).

[marcus-bcl]

Great news, sounds like this will be fixed in 2.16. Apologies, I did do a search beforehand but didn't find that existing bug.

I'll close this bug as a duplicate of opensearch-project/neural-search#813 - thanks!