azure-monitor-opentelemetry in 1.6.10 version not able to run for azure-ai-agents sdk

[Bhavyakala]

Traceback (most recent call last):
  File "C:\Users\t-kalabhavya\source\repos\azure-migrate-ai-eval-system\src\app.py", line 34, in run_async
    await self.planner_agent.create_async()
  File "C:\Users\t-kalabhavya\source\repos\azure-migrate-ai-eval-system\src\planner\planner_agent.py", line 120, in create_async
    agent = self.agent_client.create_agent(
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\t-kalabhavya\source\repos\azure-migrate-ai-eval-system\venv\Lib\site-packages\azure\ai\agents\telemetry\_ai_agents_instrumentor.py", line 1459, in inner
    return self.trace_create_agent(function, *args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\t-kalabhavya\source\repos\azure-migrate-ai-eval-system\venv\Lib\site-packages\azure\ai\agents\telemetry\_ai_agents_instrumentor.py", line 825, in trace_create_agent
    span = self.start_create_agent_span(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\t-kalabhavya\source\repos\azure-migrate-ai-eval-system\venv\Lib\site-packages\azure\ai\agents\telemetry\_ai_agents_instrumentor.py", line 746, in start_create_agent_span
    response_format=_AIAgentsInstrumentorPreview.agent_api_response_to_str(response_format),
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "C:\Users\t-kalabhavya\source\repos\azure-migrate-ai-eval-system\venv\Lib\site-packages\azure\ai\agents\telemetry\_ai_agents_instrumentor.py", line 633, in agent_api_response_to_str
    raise ValueError(f"Unknown response format {type(response_format)}")
ValueError: Unknown response format <class 'dict'>

When creating an agent using the AgentsClient SDK's create_agent function with response_format="json", the execution fails due to an unknown issue (stacktrace above). As a result, we are currently unable to use this configuration.

We have enabled via below:

import os
os.environ["AZURE_TRACING_GEN_AI_CONTENT_RECORDING_ENABLED"] = "true"  # Disabled by default

This setting enables tracing of GenAI prompts and responses, which is disabled by default.

reference:
azure-monitor-opentelemetry==1.6.10
azure-ai-agents==1.0.0

https://learn.microsoft.com/en-us/azure/ai-foundry/how-to/develop/trace-agents-sdk

[github-actions]

Thanks for the feedback! We are routing this to the appropriate team for follow-up. cc @dargilco @glharper @howieleung @jhakulin @nick863 @trangevi.

[M-Hietala]

@Bhavyakala

I was not able to reproduce this issue with code that included the following calls:

    agent = agents_client.create_agent(
        model=os.environ["MODEL_DEPLOYMENT_NAME"],
        name="my-agent",
        instructions="You are helpful agent that provides information in json format",
        response_format=AgentsResponseFormat(type="json_object")
    )


    message = agents_client.messages.create(thread_id=thread.id, role="user", content="Hello, give me a list of planets in our solar system")

Please note how the response_format parameter is passed. Also, in the instructions the response is requested in json format.

We will work to improve the documentation around this.

[wangchenghan]

@Bhavyakala
I encountered the same problem.

I modified _ai_agents_instrumentor.py. In your directory C:\Users\t-kalabhavya\source\repos\azure-migrate-ai-eval-system\venv\Lib\site-packages\azure\ai\agents\telemetry\_ai_agents_instrumentor.py at line 626, within the agent_api_response_to_str function, I added a segment of code or isinstance(response_format, dict).

Here is the complete modified function:

    @staticmethod
    def agent_api_response_to_str(response_format: Any) -> Optional[str]:
        print(response_format)
        """
        Convert response_format to string.

        :param response_format: The response format.
        :type response_format: ~azure.ai.agents._types.AgentsResponseFormatOption
        :returns: string for the response_format.
        :rtype: Optional[str]
        :raises: Value error if response_format is not of type AgentsResponseFormatOption.
        """
        if isinstance(response_format, str) or isinstance(response_format, dict) or response_format is None:
            return response_format
        if isinstance(response_format, AgentsResponseFormatMode):
            return response_format.value
        if isinstance(response_format, _models.AgentsResponseFormat):
            return response_format.type
        if isinstance(response_format, _models.ResponseFormatJsonSchemaType):
            return response_format.type
        raise ValueError(f"Unknown response format {type(response_format)}")

After making this change, my issue was resolved. I hope this helps you.