PyPi or UV support

[ChaitanyaYeole02]

Hey @tpaviot ,
I wanted to request support for Pypi or UV, as I wanted to integrated this in my current UV environment. If there is any difficulty in doing, I am open to provide support in any way possible.

[jf---]

Here’s the short version:
• pythonocc wraps OpenCascade (OCCT) and other native libs. Those C/C++ deps are hard to build and keep ABI-compatible.
• conda/conda-forge ships prebuilt, version-pinned binaries for OCCT (and friends) across macOS/Linux/Windows, so pythonocc-core “just works.”
• pip/uv are great for pure-Python (or widely-wheeled) packages, but they don’t manage system libraries. When wheels aren’t available (or mismatch your platform/toolchain), you fall back to painful local C++ builds.
• pixi keeps conda’s strong native-deps story but adds modern, project-centric workflows (lockfiles, tasks, per-project isolation, fast solver). That aligns better with an engineering/CAD/robotics stack.

⸻

Why conda is typically used for pythonocc
• Native dependency stack: pythonocc-core depends on OCCT plus graphics/tooling libs (e.g., FreeType, TBB, VTK/Qt/X11 pieces depending on build). These are system libraries, not just Python wheels.
• Binary availability: On conda-forge you get prebuilt OCCT + a pythonocc-core that’s compiled against it, with matching ABI and consistent versions.
• Cross-platform reliability: Same environment spec works on macOS/Linux/Windows without you compiling OCCT locally.
• Reproducibility: Conda’s solver and pinned packages reduce “works on my machine” issues common with native builds.

Why not (just) pip or uv for pythonocc:
• pip/uv don’t ship system libs. They install Python packages, not OCCT itself.
• Wheel gaps happen. If your platform/Python version lacks a wheel, you compile OCCT + bindings yourself (toolchain, SDKs, RPATH, headers…), which is slow and brittle.
• uv’s value is speed & isolation, not native deps. It’s a fantastic pip replacement/manager, but it won’t help resolve / install OCCT binaries.

---

Pixi might be what uv is for pip, a fast rust based implementation with the considerable benefit of reproducible builds. I think that would be worth while exploring and is perhaps and alternative route to accomplish your goal.

[ChaitanyaYeole02]

Yeah, thank you for the explanation, I have a more better idea of pythonocc and OCCT now.