Add optional rotate_pages and deskew parameters to OCR endpoint

[NataliaVillegasC]

Problem

When processing scanned PDFs with misrotated or skewed pages, the OCR output was producing incorrect text (gibberish). This is a known limitation when pages are not properly oriented before OCR is applied.

Solution

While reviewing the OCRmyPDF documentation (https://github.com/ocrmypdf/OCRmyPDF), I found two flags that address this:

• --rotate-pages: automatically detects and fixes misrotated pages
• --deskew: straightens crooked/skewed scanned pages

Both flags have been supported by OCRmyPDF for a long time and are compatible with the version used in this project.

These parameters are optional and default to false to avoid adding latency for users who don't need them.

Changes

• src/adapters/infrastructure/ocr_service_adapter.py: added rotate_pages and deskew optional parameters to process_pdf_ocr
• src/ports/services/ocr_service.py: updated abstract method signature
• src/use_cases/ocr/process_ocr_use_case.py: propagated parameters through the use case
• src/drivers/rest/app.py: added rotate_pages and deskew form parameters to the /ocr endpoint
• README.md: updated /ocr endpoint documentation with new parameters and usage examples

Testing

Tested locally with a scanned PDF with rotated pages — OCR output went from garbled text to correctly extracted content after enabling rotate_pages=true.

[gabriel-piles]

Thank you for your contribution.

We will review the code and will likely merge it next week.

Best

[gabriel-piles]

I'm going to push a few small tweaks to speed this up, hope you don't mind!

[gabriel-piles]

We are merging it

[jsurrea]

Very cool @NataliaVillegasC !