Enable sane naming of registered objects with defaults

[ashwinb]

What does this PR do?

This is a follow-up to #425. That PR allows for specifying models in the registry, but each entry needs to look like:

- identifier: ...
  provider_id: ...
  provider_resource_identifier: ...

This is headache-inducing.

The current PR makes this situation better by adopting the shape of our APIs. Namely, we need the user to only specify model-id. The rest should be optional and figured out by the Stack. You can always override it.

Here's what example ollama "full stack" registry looks like (we still need to kill or simplify shield_type crap):

models:
- model_id: Llama3.2-3B-Instruct
- model_id: Llama-Guard-3-1B
shields:
- shield_id: llama_guard
  shield_type: llama_guard

Test Plan

See test plan for #425. Re-ran it.

[yanxi0830]

Follow up: rebase and update eval_task/datasets/scoring_functions fixtures.

[ashwinb]

I fixed them now

[yanxi0830]

https://github.com/meta-llama/llama-stack-client-python/blob/85f2fb06528934dfddbef50a69ffff97af51648a/src/llama_stack_client/types/memory_bank_register_params.py#L42

Wondering if we need to regenerate OpenAPI spec with this change?

[ashwinb]

No - I don't think the actual types should change. But I will regenerate anyway @yanxi0830

[ashwinb]

Done. There are a few small changes. I am cleaning up shields anyway in the next PR so there will be more changes there.

[yanxi0830]

Wondering if the register_model should also use ModelInput?

async def register_model(
        self,
        model_id: str,
        provider_model_id: Optional[str] = None,
        provider_id: Optional[str] = None,
        metadata: Optional[Dict[str, Any]] = None,
    ) -> Model: ...

If seems like we have 3 different structures now associated with a resource?

• Model --> datastrcuture for model fields
• ModelInput --> for specifying StackRunConfig pre-registration
• register_model --> flattened args for ModelInput?

[ashwinb]

That is correct. For now, I left it as now because I think register_model()'s API looks nice as is. Adding a structure to it makes it feel unnecessarily annoying. And internally, ModelInput does get destructured and passed to register_model() so any incompatibility is immediately flagged.

[ashwinb]

Also ModelInput is purely for the Stack-server-side developer. Only necessary and used in StackRunConfig. Maybe I should think about moving it there, but it helps to keep it here to make it consistent perhaps.

[ashwinb]

I will kill @json_schema_type from these input fields for sure. They should not be in OpenAPI schema.