🚀 I'm building a tool on www.getbot.ai where users can upload PDF/Word documents.
Right now, I extract the text and send it to AI for analysis.

But here's the challenge:
📌 Many documents contain images, graphs, or charts, and I want to handle them in a smarter way.

Some approaches I'm considering:

• 📝 OCR (Tesseract, PaddleOCR, AWS Textract, Azure Read, etc.) → Extract text from images inside the docs.

• 👀 Vision models (like GPT-4o, Gemini, Claude with vision, LLaVA, Donut, etc.) → Interpret graphs/charts/images directly.

• 🔗 Hybrid workflow → First OCR the image, then pass both raw text + AI-generated description of the visual content into the analysis pipeline.

• 🗂 Embedding strategies → Store text + image captions as embeddings for semantic search and context retrieval.

💡 Questions for the community:
👉 What's the most practical way to analyze images/graphs in documents so the AI can understand them well?
👉 Any tools, libraries, or best practices you'd recommend for handling this at scale?
👉 If an entire PDF is image-based and 30–40+ pages long, what's the best approach to extract and process the content efficiently?

Thanks in advance

------------------------------
Afgan Shahguliyev
Co Founder
FutureTech Nexus
Richmond
------------------------------

Please keep questions to be about IBM API Connect here, there are other communities and forums for more general content and AI, this is just for IBM API Connect.

------------------------------
Chris Dudley
------------------------------

Original Message:
Sent: Mon August 25, 2025 10:21 PM
From: Afgan Shahguliyev
Subject: How to Handle Images & Graphs in PDF/Word AI Analysis?

🚀 I'm building a tool on www.getbot.ai where users can upload PDF/Word documents.
Right now, I extract the text and send it to AI for analysis.

But here's the challenge:
📌 Many documents contain images, graphs, or charts, and I want to handle them in a smarter way.

Some approaches I'm considering:

• 📝 OCR (Tesseract, PaddleOCR, AWS Textract, Azure Read, etc.) → Extract text from images inside the docs.

• 👀 Vision models (like GPT-4o, Gemini, Claude with vision, LLaVA, Donut, etc.) → Interpret graphs/charts/images directly.

• 🔗 Hybrid workflow → First OCR the image, then pass both raw text + AI-generated description of the visual content into the analysis pipeline.

• 🗂 Embedding strategies → Store text + image captions as embeddings for semantic search and context retrieval.

💡 Questions for the community:
👉 What's the most practical way to analyze images/graphs in documents so the AI can understand them well?
👉 Any tools, libraries, or best practices you'd recommend for handling this at scale?
👉 If an entire PDF is image-based and 30–40+ pages long, what's the best approach to extract and process the content efficiently?

Thanks in advance

------------------------------
Afgan Shahguliyev
Co Founder
FutureTech Nexus
Richmond
------------------------------