Epic Goal

The model catalog bridge will be used discover model and model server entities from a variety of sources (though note that only RHOAI MR is in scope for summit). The model catalog bridge is comprised of the following components:

• Bridge Service: Provides a REST endpoint for the backend plugin to communicate with. It should return the latest set of normalized models/model servers, based on model catalog schema. When new models/model servers are found, or entries are updated, it should push them to the model catalog plugin
• Model Catalog Data Store: The Model/Model Server metadata that the bridge fetches should be stored in a database. The bridge service and normalizers should interact with it for fetching / storing AI asset metadata. Starting implementation will be based on Git, but Bridge service + normalizer should use service provider model to ensure plugability of different database types.
• Bridge Service Normalizers: For each model/model server source (e.g. RHOAI MR) provide a service or controller that will interact with the source to fetch model and model server metadata on the cluster. It should aggregate the metadata into the model catalog schema, and store it in the data store listed above.

A current prototype implementation of the above components can be found here https://github.com/redhat-ai-dev/model-catalog-bridge 

The stories in this epic will curate this model catalog bridge ^^ to a sufficient level for our Summit '25 targets.

Why is this important?

• The model catalog bridge will be used to discover models/model servers from different sources - and will allow us to extract away the discovery logic from the backend RHDH plugin
• The normalizer(s) in the bridge will allow us to extract service specific logic (e.g. RHOAI MR, huggingface, etc) into dedicated containers for that given source

Scenarios

1. Model Catalog Bridge Service: Develop a REST endpoint for the backend plugin to communicate with and to ingest model / model servers from
2. Model Catalog Data Store: Provide a database for the model catalog bridge server & normalizers to store model catalog entities
3. Map OpenShift AI Model Registry API to our Backstage Model Catalog: Based on the model catalog schema, define a relationship between the model/model server metadata and the model catalog schema
4. Bridge Service Normalizer for RHOAI MR: Develop a service or controller that will interact with the RHOAI MR services to fetch model and model server metadata on the cluster.

Acceptance Criteria (Mandatory)

• CI - MUST be running successfully with tests automated
• Release Technical Enablement - Provide necessary release enablement details and documents.
• Bridge service is a containerized REST service, providing endpoints to fetch the latest model / model server entities, as well as handling updated model/model servers that are pushed to it by the normalizer(s).
• Model Catalog data store is based on a basic Git server, and hosts the normalized model / model server entities that are fetched from RHOAI MR
• The RHOAI MR normalizer should interact with the RHOAI MR services (in the form of a controller), and should normalize the model / model server metadata fetched from it into the standardized JSON schema that we've defined (RHDHPAI-604)

Dependencies (internal and external)

1. ...

Previous Work (Optional):

1. https://github.com/gabemontero/rhdh-ai-catalog-cli/tree/http-endpoint-for-import/

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