📊 If the answer only lives inside a chart, traditional RAG never finds it. Title: How to extract meaning from charts and tables in PDFs URL: Weaviate shows how to skip text extraction entirely and search PDF pages as images instead. Here are 3 highlights. 🔍 OCR-free multi-vector retrieval Instead of squeezing a whole page into one vector, the model keeps many vectors per image patch and scores relevance with MaxSim against the query. Charts, layouts and tables stay intact instead of being flattened into text. 🖼️ Drag-and-drop ingestion Drop PDFs into Weaviate Cloud and it renders each page as a high-res image, stores it as a BLOB, and auto-vectorizes it with a hosted module. Ingesting NVIDIA's 92-page FY26 earnings decks took about 90 seconds. 🤖 Answers with cited pages Ask "how did automotive revenue change over time" and it surfaces the exact bar chart page showing $346M to $586M (+69% YoY). The Query Agent goes further, answering with the source page images attached. The neat twist: PDFs people used to avoid because they were "mostly charts" turn out to be the ideal use case here. #RAG# #Weaviate#