[zarathustreal]

The problem with PDFs is that text isn’t necessarily text. Most RAG implementations that support them don’t do any sort of OCR or use local offline OCR implementations that have really low accuracy.