Story (Required)

As a user of the 'bac' CLI for importing new AI Models,  I want the content from Kubeflow to include the inference services endpoints

Background (Required)

During the initial drop of the CLI, the Kubeflow Model Registry did not connect KServe inferences services that it provisioned into the model registry DB. 

The Jira trackers they provided for that work have been marked resolved now.

We've received some preliminary instructions on how to obtain it, but there are still enough rough edges that an ODH (open data hub) operator (upstream of RHOAI operator)

cannot yet provision out of the box a Kubeflow Model Registry (Kubeflow and Kserve are two upstream projects for ODH).  Gabe has been working with members of the RHOAI

team to sort out how to patch ODH to get things to work.

As of Dec 11 Gabe has been able to work with the RHOAI team to stand up an ODH that sufficiently reconciles KServe inferenceservice to Kubeflow.

We also get some useful feedback on the mapping of the Kubeflow model registry logical model to our AI Model Catalog for Backstage.  Turns out their REST API is built

more around linkage than ownership.  See https://redhat-internal.slack.com/archives/C05QB0A582E/p1733515932147349 and https://miro.com/app/board/uXjVNchDNFM=/ 

Out of scope

Building catalog entries that can work with our ai templates will be a run-at exercise but not meets min for this story

Approach (Required)

• work with RHOAI to get an ODH stood up that reconciles models deployed form Model Registry into KServe InferenceServices, and the reconciles those KServer InferenceServices back to KubeFlow Inference Services, which contain the URL/Routes/Endpoints for the running models
• Update the CLI to factor in InferencesServices and how ModelArtifacts are part of the root of the Model Registry API, factoring into both feedback/tuning and versioining/deployment

Dependencies

Until ODH 2.23.0 is available, we have a set of patches needed to get ODH to properly reconcile kserve to kubeflow mr

https://redhat-internal.slack.com/archives/C05QB0A582E/p1733760430793149 for how the work in progressing.

Acceptance Criteria (Mandatory)

merged PR to our CLI repo that includes unit tests

INVEST Checklist

Dependencies identified

Blockers noted and expected delivery timelines set

Design is implementable

Acceptance criteria agreed upon

Story estimated

Legend

Unknown

Verified

Unsatisfied

Done Checklist

• Code is completed, reviewed, documented and checked in
• Unit and integration test automation have been delivered and running cleanly in continuous integration/staging/canary environment
• Continuous Delivery pipeline(s) is able to proceed with new code included
• Customer facing documentation, API docs etc. are produced/updated, reviewed and published
• Acceptance criteria are met