Epic Goal

Extend RHDH in the form of plugins and upstream changes to improve the experience managing AI assets in RHDH. Provide mechanisms to automatically import models from a variety of sources, and address limitations with modeling the AI catalog with OOTB components.

The model catalog should also provide a way to consume the model in various tools, such as pipelines, running local instances, devspaces, etc.

Why is this important?

• This will allow us to extend the model catalog beyond a simple read-only listing of model and model server artifacts
• This will allow the model catalog to hook in to other tools that need access to models

Scenarios

Extend RHDH in the form of plugins and upstream changes to improve the experience managing AI assets in RHDH:

• Provide mechanisms to automatically import models from configured instances of OpenShift AI Model Registry, 3scale, Quay
  • E.g. integrate with Kubeflow to import models from the OpenShift AI model registr
• Address limitations with modeling the AI catalog with OOTB components, and provide an experience consistent with the current Model Catalog mock-ups
• Model consumption: Models in the catalog should be consumable by the user in the form of pipelines, running their own instances, etc
• Update of the model
  • Reload of the model when the entries in the sources are changed
  • Updating the list of models when the source list has been changed
• Advanced search and filter function with query syntax defined
• This Figma [WIP shows a mockup of the user interface
• How to integrate the AI catalog with existing tools (e.g. Artifactory, Nexus, Huggingface, etc)
  • Simple approach: Can potentially document how the requirement on wrapping existing with the Open API definition that we can consume on our model catalog

Acceptance Criteria (Mandatory)

• CI - MUST be running successfully with tests automated
• Release Technical Enablement - Provide necessary release enablement details and documents.
• ...

Dependencies (internal and external)

1. ...

Previous Work (Optional):

1. ...

Open questions::

1. ...

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