Question

You are building an Azure AI Search indexing pipeline to ground an agent on a corpus of scanned technical manuals stored as multi-page PDFs in Azure Blob Storage. Each page contains embedded diagrams and photographs alongside text. You need the skillset to run OCR on the page text AND separately analyze the embedded images with a custom vision skill, so both text and image-derived content can be merged into a single searchable field per document. Which indexer configuration is required so that the embedded images are extracted and made available as inputs to the skillset?

Accepted Answer

Set the indexer's imageAction configuration to 'generateNormalizedImages' so that embedded images are cracked out and exposed via /document/normalized_images for downstream skills.

Answer

Set parsingMode to 'jsonArray' so that each embedded image is parsed as a separate JSON element and passed to the vision skill.

Answer

Enable the 'dataToExtract' setting to 'storageMetadata' so the indexer surfaces image binaries as document metadata for the skillset.

Answer

Add a DocumentExtractionSkill with default parameters, which automatically emits image regions without any indexer-level configuration.