Hello @Jack Bao 

In ODC you can build an AI Agent that reads the Word file directly and extracts exactly the information you need - text, images, and even diagram content. 

You simply upload the .docx, pass it to the agent with instructions like: “Extract all images and the text from paragraph 7”, and the agent returns: 

• The paragraph text 
• The images (Base64) 
• Any description of diagrams

Hi @Jack Bao ,

You can use a Word processing library (OpenXML, Aspose.Words, Spire.Doc) to extract paragraph text. For text inside images or diagrams, extract the images and apply OCR (Azure Vision, Tesseract). You can target specific paragraphs by index or heading before processing associated images.