Can it OCR handwritten numbers

[formlessdao]

Hi, does opendataloader recognize handwritten numbers, handwritten commas etc?
I don't see any mention of support for handwriting.

[krotname]

I would not count on reliable handwriting recognition from the current project as-is.

The README does say scanned PDFs are supported through hybrid mode OCR, for example:

pip install "opendataloader-pdf[hybrid]"
opendataloader-pdf-hybrid --port 5002 --force-ocr
opendataloader-pdf --hybrid docling-fast your-file.pdf

and the hybrid server exposes OCR engines/options such as EasyOCR by default, Tesseract, OCR language selection, and force/full-page OCR. But I do not see handwriting / HTR support called out in the docs or options, and the documented use cases are scanned PDFs, tables, formulas, charts, and structured PDF extraction.

So for printed numbers in a scanned PDF: yes, try hybrid OCR with --force-ocr and the right --ocr-lang.

For handwritten numbers or handwritten commas: I would treat that as unsupported / best-effort only. If those values matter, a safer pipeline is to run a dedicated handwriting OCR/HTR model first, then pass the cleaned text or a generated PDF layer into OpenDataLoader.