Python: Bug: Cannot use GPT-O1 Reasoning Model as Azure Chat Completion LLM Service to the kernel for agentic system

[trinaymithra]

I am trying to attach GPT-O1 Reasoning model as Azure Chat Completion LLM Service to the kernel and use it in a 'ChatCompletionAgent'. But I'm getting the following error -

kernel_v1 = Kernel()

# add LLM service to the kernel 
# GPT O1 Secrets: Agentic model secretcs
chat_service = AzureChatCompletion(
        service_id= "azure-openai",
        deployment_name=agentic_model_secrets["openai_gpt_deployment_name"],
        api_key = agentic_model_secrets["openai_api_key"],  
        api_version = agentic_model_secrets["openai_api_version"],
        endpoint = agentic_model_secrets["openai_api_base"]
)

comparison_agent = ChatCompletionAgent(
    service=chat_service,
    kernel= kernel_v1,
    name="Chat_Agent",
    instructions= """
        Answer to user questions appropriately.
    """,
    plugins=[chat_agent_plugin]
    )

ServiceResponseException: ("<class 'semantic_kernel.connectors.ai.open_ai.services.azure_chat_completion.AzureChatCompletion'> service failed to complete the prompt", BadRequestError('Error code: 400 - {'error': {'message': "Unsupported value: 'messages[0].role' does not support 'system' with this model.", 'type': 'invalid_request_error', 'param': 'messages[0].role', 'code': 'unsupported_value'}}'))

Please suggest how I can use GPT O1 Reasoning model as a service to my Chat Completion Agent.

[moonbox3]

Hello @trinaymithra, please simply configure the instruction_role as developer and it'll work without errors. Similarly shown in this sample (although this is OpenAIChatCompletion, whereas you're using AzureChatCompletion):

https://github.com/microsoft/semantic-kernel/blob/main/python/samples/concepts/reasoning/simple_reasoning.py#L59

[trinaymithra]

Still I'm getting the same error.

# GPT O1 model: Agentic Model secrets
chat_service = AzureChatCompletion(
        service_id= "azure-openai",
        deployment_name=agentic_model_secrets["openai_gpt_deployment_name"],
        api_key = agentic_model_secrets["openai_api_key"],  
        api_version = agentic_model_secrets["openai_api_version"],
        endpoint = agentic_model_secrets["openai_api_base"],
        # instruction_role = "system",
        instruction_role = "developer",
        
)

I've tried both developer and system as 'instruction_role' but still facing the same error that: Unsupported value: 'messages[0].role' does not support 'developer' with this model.

ServiceResponseException: ("<class 'semantic_kernel.connectors.ai.open_ai.services.azure_chat_completion.AzureChatCompletion'> service failed to complete the prompt", BadRequestError('Error code: 400 - {'error': {'message': "Unsupported value: 'messages[0].role' does not support 'developer' with this model.", 'type': 'invalid_request_error', 'param': 'messages[0].role', 'code': 'unsupported_value'}}'))

And, I need to use AzureChatCompletion instead of OpenAIChatCompletion as I'm working on enterprise use case and our environment is regulated via Azure

[moonbox3]

What is your api version?