Invalid schema for response_format 'ManagerSelectionResponse'

[legendgokul]

HI,

Current Setup : i have created multiple agents in AIFoundry with gpt4.1 model, each agent will have list of MCP apis attached to them under actions section

Error Info :
Error 1 :

Coordinator: {"selected_participant":"Agent1","instruction":"<instruction>","finish":false,"final_message":null}executor broke JSON is invalid: Expecting value: line 1 column 1 (char 0)

Error 2 : Coordinator/ manager agent JSON is getting duplicated as shown below

Coordinator: {"selected_participant":"<agentname>","instruction":"<instruction>","finish":false,"final_message":null}{"selected_participant":"<agentname>","instruction":"<instruction>","finish":false,"final_message":null}executor broke 1 validation error for ManagerSelectionResponse
  Invalid JSON: trailing characters at line 1 column 208 [type=json_invalid, input_value='{"selected_participant":...e,"final_message":null}', input_type=str]
   For further information visit https://errors.pydantic.dev/2.12/v/json_invalid

After which i came across below 2 link and i have tried what ever is suggest in those but the issue still exists.

1. Python: agent_framework.exceptions.ServiceResponseException: <class 'agent_framework_azure_ai._client.AzureAIClient'> service failed to complete the prompt: Error code: 400 - {'error': {'message': "Invalid schema for response_format 'ManagerSelectionResponse'.... #2681
2. Python: fix: GroupChat ManagerSelectionResponse JSON Schema for OpenAI Structured Outpu… #2750

I have tried to provide strict agent instructions to not generate improper json and also tried providing additional_chat_options as shown below :

response_schema = {
            "title": "ManagerSelectionResponse",
            "type": "object",
            "description": "Response from manager agent with speaker selection decision.",
            "properties": {
                "selected_participant": {
                    "anyOf": [{"type": "string"}, {"type": "null"}],
                    "default": None,
                    "title": "Selected Participant",
                    "description": "Name of participant to speak next (None = finish conversation)",
                },
                "instruction": {
                    "anyOf": [{"type": "string"}, {"type": "null"}],
                    "default": None,
                    "title": "Instruction",
                    "description": "Optional instruction to provide to the selected participant",
                },
                "finish": {
                    "type": "boolean",
                    "default": False,
                    "title": "Finish",
                    "description": "Whether the conversation should be completed",
                },
                "final_message": {
                    "anyOf": [{"type": "string"}, {"type": "null"}],
                    "default": None,
                    "title": "Final Message",
                    "description": "Optional text content for final message",
                },
            },
            # OpenAI strict mode requires all properties to be in required array
            "required": ["selected_participant", "instruction", "finish", "final_message"],
            "additionalProperties": False,
        }

        managerAgent = ChatAgent(            
                chat_client= AzureAIAgentClient(
                    credential=credential,
                    project_endpoint = os.getenv("AZURE_PROJECT_ENDPOINT"),
                    model_deployment_name = os.getenv("AZURE_MODEL_DEPLOYMENT"),         
                    agent_id="<agent_id>"
                ),    
                name="Coordinator" ,
                additional_chat_options={
                "response_format": {
                    "type": "json_schema",
                    "json_schema": {
                        "name": response_schema["title"],
                        "strict": True,
                        "schema": response_schema,
                    },
                },
            }                     
        )

        
        workflow = (
                GroupChatBuilder()
                .set_manager(managerAgent, display_name="Orchestrator")
                .participants([<agent list>])  
                .build()
            )

I also noticed in above issue links they have utilized AzureOpenAIChatClient but in my project i have used AzureAIAgentClient.

Note :

1. All the agents which are configured inside the workflow return basic string as output not JSON.
2. i am doing async for event in workflow.run_stream(user_message): to obtain agent streaming output.