Epic Goal

• Private/Local AI deployments like OpenShift AI are curating features similar to developers see in HuggingFace to facilitate discovery Models that those developers can leverage in their applications, though certainly at the opening of this epic, how this is going to land is still a work in progress
• Currently in our working around AI Catalogs mapping into the Backstage Catalog requires manual import to Backstage Component/Resource/API objects that are also written manually
• Hence, this epic's delves into the use of Kubeflow Model Registry as the alternative to HuggingFace, etc. for describing which Models within an OpenShift AI enabled enterprise an application developer on RHDH can use, including the models APIs, URLs for REST invocations, Keys for authenication/authorization.  The different work items in the epic will include prototyping integrations with Backstage (most likely with both the administrator and developer personas), and identify if necessary new requirements on either Backstage/DeveloperHub or Kubeflow Model Registry / ODH (as a deployer of Kubeflow Model Registry) /  OpenShift AI (as a deployer of Kubeflow Model Registry) projects

Why is this important?

• We want to ultimately provide a seamless experience for developers in Backstage discovering the up to date set of Models they are allowed to access
• We want to ultimately facilitate the steps administrators need to take to curate the set of models accessible to the various developer teams under their stewardship

Scenarios

1. ...

Acceptance Criteria (Mandatory)

• Which model registry concepts from OpenShift AI or associated projects are initially consumable by Backstage / Developer Hub
• Demo'able prototypes of the end to end flow from Model Registry the Backstage AI Model Catalog
• Productization epics/stories as needed defined on Backstage/Developer Hub
• As needed, Requirements defined on OpenShift AI related projects for Model Registry to facilitate Backstage/Developer Hub consumption

Dependencies (internal and external)

1. ...

Previous Work (Optional):

1. …

Open questions::

1. The preferred option or options that will come out of OpenShift AI still appear TBD

Done Checklist

• Acceptance criteria are met
• Non-functional properties of the Feature have been validated (such as performance, resource, UX, security or privacy aspects)
• User Journey automation is delivered
• Support and SRE teams are provided with enough skills to support the feature in production environment