Assistant: support for Sonnet 4.5 and Opus 4.1 for Bedrock

[wch]

Anthropic just announced Claude Sonnet 4.5. My understanding is that Positron will soon load Anthropic models dynamically, so it should automatically show up there when that change is merged. However, for Bedrock, the models are enumerated in a file in the Positron sources, so to add support for Sonnet 4.5 on Bedrock, Positron will need to add an entry for it.

Also, Positron currently points to Opus 4 (with both Anthropic and Bedrock), but 4.1 has been out for a while now, so we should add support for that as well.

[moa-incom]

Any progress status on when you'll have implemented the automatic load of Anthropic's models? I think a lot of us users are very keen on trying the new Sonnet 4.5 (especially since the API price is the same) :)

[wch]

I believe that functionality is part of this PR: #9345

[jmcphers]

Unfortunately today it looks like Sonnet 4.5 is only accessible via an inference profile:

You can only access Sonnet 4.5 through an inference profile. This defines which AWS Regions will process your model requests and helps manage throughput and performance.

https://aws.amazon.com/blogs/aws/introducing-claude-sonnet-4-5-in-amazon-bedrock-anthropics-most-intelligent-model-best-for-coding-and-complex-agents/

[sharon-wang]

To support inference profiles, we'd need to implement a way to call them instead of the on-demand approach we currently use for other models.

Using the Bedrock SDK directly might help us get there, as the vercel SDK may not support inference profiles directly?

These new models may be available via on-demand access in the future, once they are generally available for use.

[sharon-wang]

@timtmok observed that using the ARN works to get past the inference profile issue

from #9702 (comment)

So it does work when using the ARN that's set in the inference profile. All that had to be done is use the ARN in place of the model id. The Bedrock SDK should be able to query for available models and get the ARN. I haven't looked into when it's necessary to use the ARN as well as how to query if the model is accessible. My model queries so far have returned far more models that I would expect (looks to be all models from various providers).