Windows install for zllm-zse[cuda] fails due to Triton dependency resolution; please clarify support matrix or add platform markers

[PredictiveManish]

I’m trying to install zllm-zse[dev,cuda,training,enterprise,gguf] on Windows with Python 3.12 and CUDA 13.0.

Observed issues:

• Python 3.10 is rejected because the package requires Python >= 3.11
• On Python 3.12, uv cannot resolve zllm-zse[cuda] because triton>=2.1.0 has no matching official Windows wheels for this environment
• The project metadata lists Linux and macOS classifiers, but not Windows

Suggested improvements:

• Document that CUDA support is Linux-first / unsupported on Windows, if that is intended
• Or add platform markers so unsupported dependencies like Triton are skipped on Windows
• Or document an official Windows path if one exists (e.g. WSL2 or alternate package strategy)

Happy to work on this, if this is suggested @zyoraclub .

[zyoraclub]

Thank you for pointing the issue. We are working on this and we are upgrading the server engine to v2.0.0 with Python Kernel compiler without depending on Triton . The release will be on tuesday . you can test and let us know your feedback.

[PredictiveManish]

Okk eager for this to setup and work on other issues to make this product better.

[PredictiveManish]

@zyoraclub Any update on this? so that I can set up easily and proceed further for more bugs to find out in the codebase and fix it