Can I configure Azure Indexer to pass the entire content of PPTX and PDF documents—including text, images, tables, and graphs—through a single input field to my custom skillset for further processing with Python?

Question

Hi,

I have reports in both PPTX and PDF formats that contain text, images, tables, and various graphs. I am looking to pass the entire document content through a single input field to my custom skillset for further processing with Python code. While I am familiar with passing text and image data separately, I am wondering if it is possible to send the whole document at once using the Azure Indexer and skillset. Could you provide any guidance or insights on this?

custom skillset format:
{

  "@odata.etag": "",

  "name": "",

  "description": " ",

  "skills": [

    {

      "@odata.type": "#Microsoft.Skills.Custom.WebApiSkill",

      "name": "   ",

      "context": "/document",

      "uri": "    ",

      "httpMethod": "POST",

      "timeout": "PT1M30S",

      "batchSize": 1,

      "inputs": [

        {

          "name": " document_content",

          "source": "/document/*/data",

          "inputs": []

        }

      ],

      "outputs": [

        {

          "name": "ocr_enhanced_data",

          "targetName": "ocr_enhanced_data"

        }

      ],

      "httpHeaders": {}

    }

  ],

  "cognitiveServices": {

    "@odata.type": "   ",

    "subdomainUrl": "   "

  }

}

Answer 1

Hi there Choudhary, Mahika

Thanks for using QandA platform

I dont think the Search Indexers natively support passing the entire content of a PPTX or PDF document, through a single input field to a custom skillset. By default, the indexer extracts text and images separately, with text stored under /document/content and images under /document/normalized_images. Tables and graphs are not extracted as structured data, meaning a direct one-field input is not feasible.

maybe try modifying the custom skill to accept multiple inputs, such as both text content and images, allowing Python-based processing to merge them. Another option is preprocessing the documents before indexing using Azure Functions to convert the entire file into a Base64 string, which can then be passed as a single field to the custom skillset.

If this helps kindly accept the answer thanks much.