Azure Open AI request need a version query parameter

[witti-stephen]

In the current code of Azure Open AI client chat creation:

def chat_completions_create(self, model, messages, **kwargs):
    url = f"https://{model}.westus3.models.ai.azure.com/v1/chat/completions"
    url = f"https://{self.base_url}/chat/completions"
    if self.base_url:
        url = f"{self.base_url}/chat/completions"

This would result in a request with url:
https://openai-[region].openai.azure.com/openai/deployments/[deployment name]/chat/completions

and it would then result in a 404 resources not found error

The actual observed url should be
https://openai-[region].openai.azure.com/openai/deployments/[deployment name]/chat/completions?api-version=[version]

---

And last but not least, I like this library, it's faster than LiteLLM and as other mentioned, cleaner.

[htcd-subham]

@ksolo when we are planning to add the solution of this?

[Jamespani]

when is the next release planned for?

[wbgonz]

when is the next release planned for? We are also waiting for this update. thanks

[htcd-subham]

@ksolo any updates on this issue?

[wbgonz]

@ksolo is there any updates when this will be fixed

[wbgonz]

@ksolo unfortunately due to the lack of response we are moving away from aisuite.

[gongmingqm10]

Vote for this issue as well

[gongmingqm10]

I have added a Pull Request for solving this issue. #186

[Vahid67eb]

Incorrect URL Construction in Azure Open AI Client Chat Creation Results in 404 Error

Description:
There is an issue with the URL construction in the current Azure Open AI client chat creation code. The URL is incorrectly formed, leading to a 404 resources not found error. Below is the problematic code:

def chat_completions_create(self, model, messages, **kwargs):
    url = f"https://{model}.westus3.models.ai.azure.com/v1/chat/completions"
    url = f"https://{self.base_url}/chat/completions"
    if self.base_url:
        url = f"{self.base_url}/chat/completions"

This code results in a request with the URL:
https://openai-[region].openai.azure.com/openai/deployments/[deployment name]/chat/completions

which leads to a 404 resources not found error because it lacks the necessary api-version parameter.

Expected URL: The correct URL should include the api-version parameter, as shown below:
https://openai-[region].openai.azure.com/openai/deployments/[deployment name]/chat/completions?api-version=[version]

**Proposed Solution**: Update the URL construction to include the api-version parameter to avoid the 404 error. Here is the revised code:
def chat_completions_create(self, model, messages, **kwargs):
    api_version = kwargs.get('api_version', 'YOUR_API_VERSION_HERE')  # Ensure to replace 'YOUR_API_VERSION_HERE' with the actual version
    base_url = f"https://{model}.westus3.models.ai.azure.com/v1/chat/completions"
    if self.base_url:
        base_url = f"{self.base_url}/chat/completions"
    url = f"{base_url}?api-version={api_version}"

    # The rest of the code to send the request and handle the response
By including the api-version parameter in the URL, the request will be properly routed, resolving the 404 error.

Additional Notes: The Azure Open AI client library is notably faster than LiteLLM and, as others have mentioned, cleaner. Implementing this fix will enhance its usability even further.

Labels: bug, enhancement, Azure, AI, client, URL