Unable to instantiate and call an agent with AI Agent SDK

Question

I have created a standalone foundry project (outside of hub). I want to use AI Agent service. And I want to use it using already deployed models from a connected resource. Is it possible? Because I have connected the deployed models to the project, I can see those in the portal but have a Run error: {'code': 'invalid_engine_error', 'message': 'Failed to resolve model info for: gpt-4o'} error when using the SDK.
What is the needed configuration? I have verified that the models I am using are compatible with the region (currently Sweden Central). I have also set my environment variables OPENAI_API_VERSION, AZURE_OPENAI_ENDPOINT & AZURE_OPENAI_API_KEY.

Here is my code snippet:

async with DefaultAzureCredential() as creds:
 agents_client = AgentsClient(
endpoint=os.environ[PROJECT_ENDPOINT],
credential=creds,
 )
	async with agents_client:
		 agent = await agents_client.create_agent(
		model=os.environ['GPT_4O_DEPLOYMENT_NAME'], name="my-agent", instructions="You are helpful agent"
		 )
		print(f"Created agent, agent ID: {agent.id}")
		 thread = await agents_client.threads.create()
		print(f"Created thread, thread ID: {thread.id}")
		 message = await agents_client.messages.create(
		thread_id=thread.id, role="user", content="How to choose a washing machine"
		 )
		print(f"Created message, message ID: {message.id}")
		 run = await agents_client.runs.create(thread_id=thread.id, agent_id=agent.id)
		# Poll the run as long as run status is queued or in progress
		while run.status in ["queued", "in_progress", "requires_action"]:
		# Wait for a second
		 time.sleep(1)
		 run = await agents_client.runs.get(thread_id=thread.id, run_id=run.id)
		print(f"Run status: {run.status}")
		if run.status == "failed":
		print(f"Run error: {run.last_error}")
		await agents_client.delete_agent(agent.id)
		print("Deleted agent")
		 messages = agents_client.messages.list(
		thread_id=thread.id,
		order=ListSortOrder.ASCENDING,
		 )
		async for msg in messages:
		 last_part = msg.content[-1]
		if isinstance(last_part, MessageTextContent):
		print(f"{msg.role}: {last_part.text.value}")

Answer 1

Hi Rayane,

Thank you for contacting Q&A Forum.

To address the issue you're experiencing with the AI Agent SDK, we need to take the following steps:

1. Verify Project Endpoint: Please ensure that the project endpoint you're using is correct. This is crucial for the SDK to connect to the right resources.
2. Test the Setup: We need to test the setup on our own to verify the configuration. You can follow the example provided in the Azure AI Foundry Agents Quickstart. This will help confirm if the configuration is correct and if the issue persists.
3. Support Ticket: If the problem continues, it might be beneficial to open a support ticket with Azure. They can provide more detailed assistance and help troubleshoot any underlying issues.

Reference: https://learn.microsoft.com/en-us/azure/ai-foundry/agents/quickstart?context=%2Fazure%2Fai-foundry%2Fcontext%2Fcontext&pivots=programming-language-python-azure

If I have answered your question, please accept this answer as a token of appreciation and don't forget to give a thumbs up for "Was it helpful"!

Best Regards,

Answer 2

Hi @Rayane Laraki ,

From what I've seen at the moment, you cannot use "bring your own model" from a separate Azure OpenAI resource in Azure AI Agents Service, even if it has been connected to your Azure AI Foundry (or Azure AI Foundry project): the model deployment you are using must be deployed within the Azure AI Foundry resource of your agent.

See also the question I have posted in the Azure AI Foundry GitHub discussions about the same topic: https://github.com/orgs/azure-ai-foundry/discussions/59

Answer 3

Hi Nicolas,

I got the same error. i fixed that deployin a model in AI foundry, not in open ai. After deploying, you can in the platform assign it to your agent.

BR/