CODEWITHSUNDEEP
azureazure-functionsazure-blob-storageazure-cognitive-search
hougumnaah
hougumnaah

Reputation: 45

Custom WebApi Skill does not populate index

I am new to Azure Search, and I am currently setting up an environment to index images based on certain details. I have set up an index that has, among other fields, a collection of complex fields I named details.

To populate the details field, I have set up an Azure Function as a WebApiSkill on my indexer, that receives the uri as input and returns the details.

According to the example from the documentation, the function receives a application/json POST request as follows:

{
    "values": [
        {
            "recordId": "1",
            "data": {
                "uri": "..."
            }
        },
        ...
    ]
}

..and promptly returns an OkObjectResult with the corresponding json response:

{
    "values": [
        {
            "recordId": "1",
            "data": {
                "details": [
                    {
                        "detailName": "...",
                        "detailRatio": "..."
                    }, {
                        "detailName": "...",
                        "detailRatio": "..."
                    }
                ]
            }
        },
        ...
    ]
}

My problem is, even though I followed the documentation on how to setup a custom skill for an indexer's skillset, and even though all of the above seem properly structured, the details on my index's images remain unpopulated, and it just shows as an empty list. Here are the results of an empty search on my index:

{
    "@odata.context": "https://{search}.search.windows.net/indexes('{index}')/$metadata#docs(*)",
    "value": [
        {
            "@search.score": 1,
            "id": "aHR0CHM6Ly9yZWxhdGl...",
            "uri": "{link}",
            "details": []
        }
}

By following the Azure Function's execution through console logs, it's possible to see that the indexer is indeed executing it and receiving the result, so I assume it's not a connection issue or the function's code's problem.

I would like to know if this is a common occurence? And if not, what steps should I take to try and figure out why the indexer refuses to properly populate the indexes?

Here are some code snippets from my indexes/indexer/skillset. I apologize for any confusion here, as some of these had to be made by code (C#) while others were made in the Azure Portal or by a POST request. To my knowledge, all of these boil down to the same, but I am currently in the process of converting all of these to POST requests.

Index:

{
    "id": "...",
    "uri": "...",
    "details": [
        {
            "detailName": "...",
            "detailRatio": "..."
        }, {
            "detailName": "...",
            "detailRatio": "..."
        }
    ]
}

Skillset:

{
  "name": "...",
  "description": "...",
  "skills":
  [
    {
      "@odata.type": "#Microsoft.Skills.Custom.WebApiSkill",
      "uri": "https://{azure-function}.azurewebsites.net/api/{function}?code={code}",
      "context": "/document",
      "inputs": [
        {
          "name": "uri",
          "source": "/document/uri"
        }
      ],
      "outputs": [
        {
          "name": "details",
          "targetName": "details"
        }
      ]
    }
  ]
}

Indexer:

new Indexer(
                name: indexerName,
                dataSourceName: dataSourceName,
                targetIndexName: indexName,
                skillsetName: skillsetName,
                fieldMappings: new List<FieldMapping>
                {
                    new FieldMapping
                    {
                        SourceFieldName = "metadata_storage_path",
                        TargetFieldName = "uri"
                    }
                },
                schedule: new IndexingSchedule {
                    Interval = new TimeSpan(0, 10, 0)
                },
                parameters: new IndexingParameters
                {
                    Configuration = new Dictionary<string, object>
                    {
                        ["imageAction"] = "generateNormalizedImages"
                    }
                }
            );

Upvotes: 1

Views: 340

Answers (1)

8163264128
8163264128

Reputation: 787

Your indexer definition is missing outputFieldMappings to specify where the data for details should come from. See https://learn.microsoft.com/en-us/rest/api/searchservice/create-indexer#outputfieldmappings

The context of your skills is "/document" and targetName is "details", so the result will be at "/document/details".

"outputFieldMappings": [
  {
    "sourceFieldName": "/document/details",
    "targetFieldName": "details"
  }
]

Upvotes: 2

Related Questions