[ANNOUNCE] Kubeflow SDK 0.4.0 is released!

[Andrey Velichkevich]

Hi Kubeflow Community,

We’re excited to announce the release of Kubeflow SDK 0.4 🎉

pip install -U kubeflow

This release marks a big step toward a vision we’ve been working on: a unified Python SDK for running any AI workloads at any scale – without the usual Kubernetes infrastructure complexity.

It introduces two new clients:

• SparkClient for scalable AI data processing
  
• ModelRegistryClient for registering, retrieving, and versioning AI models

With these additions, the Kubeflow SDK moves closer to supporting the entire AI lifecycle: data processing → training → optimization → model management

Full details are available in:

• LinkedIn announcement 
  
• Blog post

If you are at KubeCon + CloudNativeCon EU 2026 stop by the Kubeflow booth for a deep dive, or attend the demo at the project pavilion stage!

🙌 Huge thanks to all contributors who made this release possible.

Regards,
Andrey