Views:

Applies to Product - Microsoft Copilot Studio


What’s happening?
Customers are experiencing issues with the Azure OpenAI connection when deploying AI models into Copilot Studio. The generative answer feature is not functioning as expected, resulting in empty responses or loss of connection to the Azure OpenAI service.
 

Reason:
This may arise due to the deployment configuration for the Azure AI model not including the necessary settings to maintain the connection to the Azure OpenAI service. Additionally, it has been noted that deploying to Copilot Studio is only available in certain regions, and the feature may be in preview, which can lead to connectivity issues.
 

Resolution:

  • Ensure that the deployment configuration for the Azure AI model includes the necessary settings to maintain the connection to the Azure OpenAI service. Verify that the correct resource and endpoint are being used.
  • Confirm that the Azure OpenAI resource is available and properly configured in the Azure subscription. Resource limitations or misconfigurations can cause connectivity issues.
  • Review the authentication settings for both the Azure AI model and the Azure OpenAI service to ensure that the appropriate permissions and credentials are in place.
  • If the connection issue persists, access the Power Platform admin center from an in-private browser and monitor the behavior.
  • Follow the steps outlined in the documentation for connecting data sources to Copilot Studio, ensuring that data is added in the Azure OpenAI playground before attempting to connect to a copilot in Copilot Studio. Refer to the documentation for "Generative answers with Azure OpenAI on your data (preview) - Microsoft Copilot Studio" for further guidance.
If the issue remains unresolved after following these steps, please provide detailed screenshots of the model theme and model-driven configurations, along with confirmation of the time zone, to assist in further troubleshooting.