Enhance flexibility by supporting deployment_id for Azure OpenAI model switching

[ZhenhangTung]

Is your feature request related to a problem? Please describe.
Currently, the Azure OpenAI SDK requires setting a full deployment URI, which makes switching models for different tasks cumbersome and affects cost efficiency, especially when using more expensive models like ChatGPT-4 only for complex tasks.

Describe the solution you'd like
I suggest allowing deployment_id specification when initializing a new client, similar to other SDKs. This would enable easy model switching, improving flexibility and cost management.

Describe alternatives you've considered
An alternative is overwriting the uri_base during client setup, but it's less convenient than simply changing a deployment_id.

Additional context
This feature is important for applications that need different models for efficiency and cost-effectiveness, and for developers experimenting with multiple models.

I am very keen on seeing this feature implemented and am willing to offer my help and support in making it happen. If needed, I can contribute to development, testing, or documentation.

[washu]

you can monkeypath this easily
module OpenAI
class Client
def self.azure_deployment(model_name,is_azure)
# I use encrypted credentials store to fetch the correct uri base
uri_base = ... lookup your base id
OpenAI::Client.new(uri_base: isAzure ? uri_base + "openai/deployments/#{model_name}" : nil)
end
end
end