CODEWITHSUNDEEP
azuresearchmetadataazure-cognitive-search
Axxeption
Axxeption

Reputation: 293

Is it possible to use metadata as input for custom skill in azure search?

I started working with Azure search a couple of months ago but I have an issue with the metadata of blobfiles.

I need the metadata of a file (coming from Azure Blob) to use it in my customskill. (More specific I need the URL of the blobfile where it's stored).

To do this I need it in my skillset, I would do something like in this image. But that's not possible because the source has to start with /document? If I do "/document/metadata_storage_path/" as "Source" I got a null value in the end?

Is there a way to get the metadata of a file as input to use it further on?

Thanks in advance!

Upvotes: 0

Views: 799

Answers (2)

Axxeption
Axxeption

Reputation: 293

I discovered why it (and the solutions) were not working for me. I hope I can help others that encounter this issue.

The syntax mentioned above by Sophiac was correct. So in my case I used "metadata_storage_path" as input in the skillset:

{
      "@odata.type": "#Microsoft.Skills.Custom.WebApiSkill",
      "description": "Our new substring custom skill",
      "uri": "https://customskillsubstring.azurewebsites.net/api/Translate?code=OkzL7G3wX----jCqQylUyJJPaggSaFQCaQ==",
      "batchSize":1,
      "context": "/document",
      "inputs": [
        {
          "name": "text", "source": "/document/metadata_storage_path"
        }
      ],
      "outputs": [
        {
          "name": "text",
          "targetName": "metadata_storage_path_wathever"
        }
      ]
    }

The issue was in the indexer. I mapped in the fieldmapping the "metadata_storage_path" to something else (in my case "blob_uri"). The problem is this is not really a mapping but more like a replacement. So the "metadata_storage_path" was empty in the skillset because it was already replaced.

But if I use "blob_uri" it is working. Solution is that you can map one input to more than one thing in the indexer: 

"fieldMappings" : [
        {
          "sourceFieldName" : "metadata_storage_name",
          "targetFieldName" : "id",
          "mappingFunction" :
            { "name" : "base64Encode" }
        },
        {
          "sourceFieldName" : "content",
          "targetFieldName" : "content"
        },
        {
          "sourceFieldName" : "metadata_storage_path",
          "targetFieldName" : "blob_uri"
        },
        {
          "sourceFieldName" : "metadata_storage_path",
          "targetFieldName" : "metadata_storage_path"
        }
   ],

Now I can use the "blob_uri" and the "metadata_storage_path" as an input for my customskill.

Upvotes: 2

Sophiac
Sophiac

Reputation: 173

I think your source path would have to be "/document/metadata_storage_path" without the extra "/" at the end. With the extra "/", the source path is interpreted to be a directory within metadata_storage_path that has the name of "" (empty string).

Upvotes: 0

Related Questions