Context
Today, the existing AzureOpenAITextToAudioService service and the new one try to use model id supplied via constructor, then if that is not provided tries to find it in prompt execution settings and if there's nothing there the deployment name is used.
ToDo
Revise the logic of both existing and the new services:
- Ensure that prioritizing the model id supplied in the constructor over the more "dynamic" model id specified in the prompt execution settings (which can be changed per operation rather than per instantiation) is the correct behavior.
- Verify whether the model id parameter/property is still required or relevant, given that the new Azure.AI.OpenAI SDK v2 supports either a deployment name or a model id, but not both at the same time, as the HTTP version of the existing
AzureOpenAITextToAudioService does.
Context
Today, the existing
AzureOpenAITextToAudioServiceservice and the new one try to use model id supplied via constructor, then if that is not provided tries to find it in prompt execution settings and if there's nothing there the deployment name is used.ToDo
Revise the logic of both existing and the new services:
AzureOpenAITextToAudioServicedoes.