Hub docs don't mention/relate the new kubeflow.hub.ModelRegistryClient (Kubeflow SDK v0.4.0) alongside the legacy model-registry client

[Ishan-AI-coder]

What happened

The Hub documentation (Getting Started, Python Client reference) currently documents only the standalone Model Registry Python client:

from model_registry import ModelRegistry

However, since Kubeflow SDK v0.4.0, there is also a unified Hub client available under the Kubeflow SDK:

from kubeflow.hub import ModelRegistryClient

This client is documented at the Kubeflow SDK documentation and uses a different import path and constructor signature.

The current website does not mention that this newer client exists or explain how it relates to the legacy standalone client, which can be confusing for users who are starting with Kubeflow Hub today.

Additionally, a few pages under /docs/components/hub/reference/ (for example, python-client.md and architecture.md) still refer to "Model Registry Python Client" in the body text after the Hub rename, so the terminology is slightly inconsistent.

Suggested fix

• Add a short note or callout to the Getting Started guide mentioning the kubeflow.hub.ModelRegistryClient available in the Kubeflow SDK.
• Add a similar note to the Python Client reference page explaining that there are currently two client libraries, with a brief explanation of how they differ (or which one is recommended going forward).
• Perform a small terminology consistency pass across the Hub documentation, replacing outdated references to "Model Registry Python Client" where appropriate.

Additional context

I'd be happy to submit a PR for this if a maintainer can clarify the intended direction:

• Should the standalone model_registry client now be considered the legacy/deprecated client?
• Or are both clients expected to remain first-class supported interfaces?

[Ishan-AI-coder]

/cc @franciscojavierarceo, @thesuperzapper, @andreyvelich

Could you please provide some guidance on the intended direction for this issue? That will help me proceed with the implementation. Thanks!

[christian-heusel]

@Ishan-AI-coder this is a topic for @kubeflow/kubeflow-hub-team and not neccessarily the Kubeflow Steering Committe Members 🤗
@jonburdo did a lot of the refactor in #4364, but I guess there are still a few places left 😄

[Ishan-AI-coder]

Thanks for the clarification!@christian-heusel I'll reach out to the Kubeflow Hub Team instead. I appreciate the pointer to jonburdo and PR #4364—I'll review it to better understand the remaining changes.

[christian-heusel]

@Ishan-AI-coder lets not tag people too often, they already got a ping with my initial message 🤗

[Ishan-AI-coder]

Got it, thanks! I'll avoid additional @mentions and let the discussion proceed naturally.