# Kagi Assistant does not send files to models properly > Source: > Published: 2026-10-16 19:29:29+00:00 I wan't to OCR PDFs to Markdown TEXT using a model of my choice that supports ingesting PDFs. How do I do this? Output of the "Read source" step: No relevant content could be extracted: the attached PDF was unavailable for processing, so its text cannot be quoted verbatim. Output of the "Contemplating" step: Analyzing File Errors I'm currently investigating the failure to load a specific PDF file, identified as uploads/foobar.pdf. The error message points to an issue with the document's size not meeting a minimum requirement. This seems to be the primary cause for the read failure. I plan to check the file's dimensions. Final output: [unreadable: The attached file foobar.pdf could not be processed by the document reader ("Min document size not reached"), so its contents are unavailable and cannot be transcribed.] According to [https://kagifeedback.org/d/6801-gemini-use-document-processing-for-much-better-pdf-handling](https://kagifeedback.org/d/6801-gemini-use-document-processing-for-much-better-pdf-handling) the problem I'm describing shouldn't exist anymore... but it does! What should have happened is that Kagi Assistant sends the raw file to the model.