Chapter 1. Manage and govern model catalog sources in the OpenShift AI dashboard


As an OpenShift AI administrator, you can manage and govern model catalog sources to control which models are available to data scientists and AI engineers in your organization. Model catalog sources make models available from different providers that users can discover, register, and deploy.

The OpenShift AI dashboard includes model catalog settings screens where you can add, edit, preview, and delete catalog sources. You can configure model visibility settings to allow or disallow specific models for governance and compliance purposes. You can also enable or disable a model catalog source.

1.1. View model catalog sources

In the OpenShift AI dashboard, under Settings Model resources and operations Model catalog settings, you can view all configured model catalog sources and their status. Each catalog source displays the following information:

  • Source name: The name of the source to be displayed in the catalog. The default sources are Red Hat AI, Red Hat AI validated, and Other.
  • Organization: For Hugging Face sources only, the organization that provides the models, for example, meta-llama. Only models in this organization are included in the catalog.
  • Model visibility: Indicates whether all models from a source display in the catalog or if models are filtered based on visibility settings.
  • Source type: The catalog source type, either Hugging Face repository or YAML file.
  • Enable: You can enable a source to make its models available in the catalog for users in your organization.
  • Validation status: When a non-default source is enabled, the validation status indicates whether the source has successfully connected to the catalog.

You can filter catalog sources by source name, organization, model visibility, source type, and enabled status to quickly find the sources that you want to manage.

1.2. Enable or disable catalog sources

When you add a new catalog source, you can keep it disabled while you configure the source details and model visibility settings. After you finish configuring the source, you can enable it to make the models available in the catalog for users.

When you enable a catalog source, OpenShift AI validates the connection to ensure that the source is accessible. The validation status updates to show when the connection is starting, successful, or failed.

You can disable a catalog source at any time to remove its models from the catalog without deleting the source configuration.

For governance and compliance, you can configure model visibility settings to control which models from a catalog source are available to users in your organization. This capability is essential for highly regulated organizations, where you must ensure that specific third-party models are not exposed to end users due to intellectual property concerns, geopolitical risks, or compliance requirements.

You can use model visibility settings to include or exclude specific models by using pattern matching and wildcards. OpenShift AI applies included patterns first to select models from the catalog source, and then applies excluded patterns to filter out specific models from that set.

1.4. Limitations for Hugging Face model sources

The following limitations apply for Hugging Face repository sources:

  • Public non-gated models only: This feature does not support gated or private models that require API keys or secret management.
  • Connectivity requirements: OpenShift AI does not support deployment of this feature in disconnected environments. The cluster must have external network access to Hugging Face.
  • Model validation: Red Hat does not validate or guarantee the security of third-party models. You are responsible for vetting all external content.
  • Performance: Deployment times depend on model size and network speed, both of which might increase initial startup latency.
Red Hat logoGithubredditYoutubeTwitter

Learn

Try, buy, & sell

Communities

About Red Hat

We deliver hardened solutions that make it easier for enterprises to work across platforms and environments, from the core datacenter to the network edge.

Making open source more inclusive

Red Hat is committed to replacing problematic language in our code, documentation, and web properties. For more details, see the Red Hat Blog.

About Red Hat Documentation

Legal Notice

Theme

© 2026 Red Hat
Back to top