Fix docs related to Azure OpenAI's "model"

[sethupavan12]

Why are these changes needed?

The current documentation for using Autogen with Azure OpenAI contains an error. It states that the model parameter accepts values like gpt-4 or gpt-3.5-turbo. However, in practice, this parameter should be set to the deployment name, which can be any user-defined string. This discrepancy in the documentation wasted a lot of time for me. Hopefully, this can save some time for others.

Related issue number

Docs concern raised in #822

Checks

• I've included any doc changes needed for https://microsoft.github.io/autogen/. See https://microsoft.github.io/autogen/docs/Contribute#documentation to build and test documentation locally.
• I've added tests (if relevant) corresponding to the changes introduced in this PR.
• I've made sure all auto checks have passed.
• I've tested that using model name like gpt-4 doesn't work
• I've tested that using deployment name instead of model works

[jackgerrits]

It uses the model as the deployment if the deployment is not specified. See the logic here:

autogen/autogen/oai/client.py

Line 397 in 94bff2d

def _configure_azure_openai(self, config: Dict[str, Any], openai_config: Dict[str, Any]) -> None:

I do agree this could be made clearer. Given the above logic do you think the proposed change should be updated?

[sethupavan12]

It uses the model as the deployment if the deployment is not specified. See the logic here:

autogen/autogen/oai/client.py

Line 397 in 94bff2d

def _configure_azure_openai(self, config: Dict[str, Any], openai_config: Dict[str, Any]) -> None:

I do agree this could be made clearer. Given the above logic do you think the proposed change should be updated?

Shouldn't the logic raise an error if the deployment name is not specified as it is a mandatory field?
Could you explain why model is used in deployment's absence?

[jackgerrits]

@sonichi can you comment?

As I understand it - often deployments are named the same as model names so this was simply a shortcut for this case. It may be too "magic" and leads to this sort of confusion

[BeibinLi]

LGTM. Good catch!