Implementation for AI model artifact processor

[chlins]

Background

https://github.com/CloudNativeAI/model-spec

The model spec extends the OCI image spec for better building and serving the AI models to the OCI Artifact, since Harbor is the OCI Artifact registry, so we propose that Harbor adopts the model specification standards to ensure consistency and compatibility with AI model-related workflows. Additionally, we request the development of an experimental feature to enable the display of foundational AI model metadata, as previous design in the Harbor codebase, we only need to implement an AI model processor to handle related logics and some UI works for better user interaction experience.

[reasonerjt]

What does kind/experimental mean? If this spec is not GAed will we merge the change to main branch?

If you want to display customized icon or metadata, could you follow this flow:
https://github.com/goharbor/community/blob/main/proposals/enhanced-default-processor.md

[chlins]

I believe that AI model may require more intricate logic in the future, and the current enhanced default processor is only capable of customizing icons or metadata, which falls short of handling certain extraction tasks such as helm charts. Therefore, I propose that we develop a new processor specifically for the AI model, which will offer enhanced customization and flexibility. At the very least, if we intend to incorporate this modification, we must publish one release version of model specification.

For the kind/experimental label, I am adding it to indicate a feature that we can consider as experimental.

[reasonerjt]

@chlins
A general gap is that there's no such concept as "experimental feature" or "feature flag" in Harbor.
Would you mind write up a short proposal to better explain why the customization mechanism is not enough in v2.13 so that you need a new processor in v2.13.0? You may also propose a way to optionally enable this processor, which I assume will be disabled by default given you want it to be experimental. And we can review and discuss and make final call before the feature freeze date.

[chlins]

If we don't currently have a concept like experimental feature on harbor, I think it's also possible to not have this label, because adding a new model type processor is not going to have any impact on the existing codebase or existing functionality, but I agree that we could have a short proposal to explain why we need a new processor to better handle model parsing.