fix: Supported model names

[ZhongpinWang]

Context

AI/gen-ai-hub-sdk-js-backlog#113.

Model name autocompletion

• for openai in foundation-models
• for orchestration

Model types are defined centrally in core and then separately imported into these two packages.

Definition of Done

• Code is tested (Unit, E2E)
• Error handling created / updated & covered by the tests above
• Documentation updated
• (Optional) Aligned changes with the Java SDK
• (Optional) Release notes updated

[MatKuhr]

Maybe we can define the constants only once and use the union of LLM Access + orchestration. And then use Omit<.> within the specific clients?

While there will be exceptions, probably most models will be the same across both, so we would only have to omit a few models, instead of duplicating a lot..

[ZhongpinWang]

@MatKuhr @deekshas8 Thanks for your feedback! Still a bit hesitating if excluding models would be a good idea as per default, as per default, adding a new model in a reused model type file and then exclude is a bit error-prone as we might simply forget to exclude it. But anyways since it is now quite simple to do so, I weight reusability a bit higher than maintainability, and refactored accordingly.

[MatKuhr]

we might simply forget to exclude it.

Equally, we might forget to add a model in two places, i.e. by accident only adding it in one. For now we can assume 9/10 models will be available in both LLM access & orchestration, so I'd rather take the chance on exclusion 😉

[deekshas8]

LGTM