Epic Goal

The AI Catalog will utilize the out of the box Backstage Software Catalog, API and TechDocs to store and share searchable information about a company’s AI services, these information items should include:

• The LLM data files tracked by the customer (e.g. GGUF, safetensor files, etc.), modeled in the RHDH catalog as Resources.
• Model APIs, modeled in RHDH as API entities and readable in the API viewer in RHDH if using compatible specifications (like Swagger, OpenAPI, gRPC, etc).
• Model servers, modeled in the catalog as Components These may be linked to APIs and/or Resources using the entity graph in RHDH.
• Model information and usage instructions, modeled in RHDH as TechDocs.

The AI catalog should support basic search and filter function with limited number of fields, based on what’s supported OOTB  in RHDH.

Why is this important?

• This will define how model resources are modeled in Backstage, and will provide the basis for all subsequent work done on the model catalog
• This will allow an AI catalog to be modeled using native Backstage types and resources, without the need for plugins.

Scenarios

1. Backstage API implementation: The user should be able to model their AI catalog (model, model servers, and APIs) using native Backstage types
2. Model Documentation: The AI developer should be able to access information on: Model information, purpose, license, and type
3. Model Usage: The AI developer should be able to access information on: How to access the model server, download, and usage restrictions
4. Testing: The catalog structure should be validated against a real world scenario

Acceptance Criteria (Mandatory)

1. Entire Model Catalog API implemented: RHDP-1014: Model Catalog API
2. Model and Model server information documented in the form of techdocs
3. Implementation validated against real-world scenarios

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