I've been trying to use Elicit to analyze some lengthy AWS Well-Architected review PDFs and Azure architecture whitepapers, but the uploader consistently times out on files over ~50MB. Given that a single month's cost anomaly report from our CUR can easily hit 100MB+, this is a blocker for my FinOps workflow.
Has anyone found a reliable workaround? I'm considering:
* Pre-splitting PDFs by section (though this loses document context)
* Using a local tool to compress the PDFs first
* A script to extract and upload just the text layer
The ideal would be if Elicit's API supported chunked uploads, but I don't see that documented. For now, my process is stalled. I even tried converting a 75MB PDF to images and then OCR-ing, but that seems like a convoluted step.
If you've tackled this, what worked? Bonus points if you have a quick Python or bash snippet to pre-process files for Elicit. Something like:
```python
import PyPDF2
from pathlib import Path
def split_pdf_for_elicit(file_path, max_size_mb=40):
# Basic size-based chunking logic
pass
```