Hello
I was testing glmocr, and one of my verification scripts flagged a licensing issue related to a mandatory dependency.
The project declares Apache-2.0 but has pymupdf>=1.24.0 as a mandatory dependency, and PyMuPDF is licensed under GNU AGPL v3.0. These two licenses are incompatible: the AGPL copyleft propagates to any work that includes it, which means glmocr effective license ends up being AGPL-3.0 in practice, not Apache-2.0.
The fix looks straightforward. Either update the project license to reflect the effective AGPL-3.0 implications of the current dependency tree, or replace pymupdf with an Apache-2.0-compatible alternative such as pypdfium2 in the base installation. Since pypdfium2 is already present in the selfhosted and all extras, the second option seems practical for this use case.
This same issue has come up and been resolved in similar projects: browser-use (browser-use/browser-use#2610), doctr (mindee/doctr#486) and gpt-researcher (assafelovic/gpt-researcher#1063).
Hello
I was testing glmocr, and one of my verification scripts flagged a licensing issue related to a mandatory dependency.
The project declares
Apache-2.0but haspymupdf>=1.24.0as a mandatory dependency, and PyMuPDF is licensed under GNU AGPL v3.0. These two licenses are incompatible: the AGPL copyleft propagates to any work that includes it, which means glmocr effective license ends up being AGPL-3.0 in practice, not Apache-2.0.The fix looks straightforward. Either update the project license to reflect the effective AGPL-3.0 implications of the current dependency tree, or replace pymupdf with an Apache-2.0-compatible alternative such as pypdfium2 in the base installation. Since pypdfium2 is already present in the selfhosted and all extras, the second option seems practical for this use case.
This same issue has come up and been resolved in similar projects: browser-use (browser-use/browser-use#2610), doctr (mindee/doctr#486) and gpt-researcher (assafelovic/gpt-researcher#1063).