support numpy >= 2.0 since it errors out on numpy >=2.0

[jj-github-jj]

Requested feature

numpy > 2.0 is used on modern systems but this module requires numpy < 2 to not errror out

...

Alternatives

...

[hnc-jglee]

Thank you for reporting this issue.

Looking at the current state of the project, numpy >= 2.0 is already supported. Our uv.lock resolves to numpy 2.2.6 (Python < 3.11) and numpy 2.4.2 (Python >= 3.11), and our upstream dependencies (docling >= 2.0.0, easyocr 1.7.2, etc.) are all compatible with numpy 2.x.

This project does not declare any direct numpy version constraint — numpy is pulled in as a transitive dependency through the hybrid optional dependencies (docling[easyocr]).

If you're encountering an error related to numpy >= 2.0, it may be due to one of the following:

1. Stale installation — An older cached version of a dependency (e.g., an older docling or easyocr) may still pin numpy < 2. Try reinstalling from scratch:

   pip install --upgrade --force-reinstall opendataloader-pdf[hybrid]

2. Environment conflict — Another package in your environment may be constraining numpy < 2. You can check with:

   pip check

3. Specific error message — Could you share the exact error traceback you're seeing? That would help us pinpoint whether this is a dependency resolution issue or a runtime incompatibility.

If you can provide:

• The Python version you're using
• The exact error message/traceback
• How you installed the package (pip, uv, etc.)

…we'll be able to help narrow down the cause.

[bundolee]

Closing this issue due to inactivity — no response in over a month after our request for additional details.

To reiterate: as of the current main branch, numpy >= 2.0 is already supported. Our lockfile resolves to numpy 2.2.6 / 2.4.2, and there is no direct numpy version constraint in this project.

If you're still encountering this issue on a recent release, please feel free to reopen with:

• The Python version
• The exact error message/traceback
• How you installed the package (and the resolved versions of numpy, docling, easyocr)

We'd be happy to investigate further with that information.

---

🤖 AI Issue Fix — status: closed-stale

Reporter claimed numpy < 2 was forced, but the project already supports numpy 2.x (uv.lock: 2.2.6 / 2.4.2; no direct numpy constraint). Maintainer requested traceback + version info on 2026-03-23; no response in 38 days. Closing as stale; reopen with reproduction details welcome.