Use neural search with AI Chat searching

[adamjarling]

What does this do?

Updates the OpenSearch query object to use a hybrid search (using neural), when a user is searching via AI Chat.

How to test?

Note it's a little tough to test UI search results against limited staging data

• Update your local miscellany repo for env variables needed for hybrid search
• Perform a regular and Chat AI search.
• Inspect the request body, and notice when using AI Chat search (if a search term is provided), the request payload should resemble something like:
• The Search results should visually appear as expected, but the paginated results will be calculated differently by the API

{
   ...
    "query": {
        "hybrid": {
            "queries": [
                {
                    "bool": {
                        "must": [
                            {
                                "query_string": {
                                    "fields": [
                                        "title^5",
                                        "all_text",
                                        "all_controlled_labels",
                                        "all_ids"
                                    ],
                                    "query": "Do we have music by Mozart?"
                                }
                            }
                        ]
                    }
                },
                {
                    "neural": {
                        "embedding": {
                            "k": 40,
                            "model_id": NEXT_PUBLIC_OPENSEARCH_MODEL_ID,
                            "query_text": "Do we have music by Mozart?"
                        }
                    }
                }
            ]
        }
    },
    "size": 40
}

[bmquinn]

The query looks great in my dev env. It's really awesome seeing the hybrid query results coming in while the streaming response is being generated! 😄