[Bug]: Cannot extract summary using reflection_with_llm: Error code: 400

[Alaya-Con]

Describe the bug

I use the agents chat. Both the user_proxy and assistant have nested chats with summary "reflection_with_llm".

UserWarning: Cannot extract summary using reflection_with_llm: Error code: 400
{'type': 'literal_error', 'loc': ('body', 'messages', 2, 'typed-dict', 'role'), 'msg': "Input should be 'function'", 'input': 'user', 'ctx': {'expected': "'function'"}}, {'type': 'extra_forbidden', 'loc': ('body', 'messages', 2, 'typed-dict', 'tool_calls'), 'msg': 'Extra inputs are not permitted', 'input': []}, {'type': 'extra_forbidden', 'loc': ('body', 'messages', 2, 'typed-dict', 'tool_calls'), 'msg': 'Extra inputs are not permitted', 'input': []}]', 'type': 'BadRequestError', 'param': None, 'code': 400}. Using an empty str as summary.

Code block:

def writing_message(recipient, messages, sender, config):
    return f"{recipient.chat_messages_for_summary(sender)[-1]['content']}"

nested_chats1 = [
    {
        "recipient": coder,
        "message": writing_message,
        "clear_history": True,
        "summary_method": "reflection_with_llm",
        "summary_args": {"summary_prompt": '''
        Return the original task and the final improved code block to solve the task.
        '''},
        "max_turns": 1
    },
]
nested_chats2 = [
    {
        "recipient": executor,
        "summary_method": "last_msg",
        "max_turns": 1
    },
    {
        "recipient": tester,
        "message": writing_message,
        "summary_method": "reflection_with_llm",
        "summary_args": {"summary_prompt": '''
        Return the exitcode, Code output print and the result analysis. 
        If the result indicates there is an error, then reply "Please improve the code." in the end. If the result is correct, then reply "TERMINATE" in the end.
        '''},
        "max_turns": 1
    },

]

user_proxy = autogen.UserProxyAgent(
    name="user_proxy",
    human_input_mode="NEVER",
    max_consecutive_auto_reply=3,
    is_termination_msg=lambda x: x.get("content", "").find("TERMINATE") >= 0,
    # is_termination_msg=lambda x: len(x.get("content", "").rstrip()) < 2,
    code_execution_config=False,
    # code_execution_config={
    #     # the executor to run the generated code
    #     "executor": LocalCommandLineCodeExecutor(work_dir="local_coding"),
    # },
)

assistant = autogen.AssistantAgent(
    name="assistant",
    llm_config={
        "cache_seed": 41,  # seed for caching and reproducibility
        "config_list": config_list,  # a list of OpenAI API configurations
        "temperature": 0,  # temperature for sampling
    },
    is_termination_msg=lambda x: x.get("content", "").find("TERMINATE") >= 0,
)

user_proxy.register_nested_chats(
    nested_chats2,
    trigger=assistant,
)

assistant.register_nested_chats(
    nested_chats1,
    trigger=user_proxy,
)

chat_res = user_proxy.initiate_chat(
        assistant,
        message=task,
        max_turns=2,
        summary_method="reflection_with_llm",
        summary_args={"summary_prompt": "Return the final code block to solve the task, whose test result is correct."},
    )

Steps to reproduce

No response

Model Used

llama3-70b-instruct
Use latest vllm:
CUDA_VISIBLE_DEVICES=0,1,2,3 python -m vllm.entrypoints.openai.api_server --model ./Meta-Llama-3___1-70B-Instruct --host 0.0.0.0 --port 50051 --served-model-name llama3-70b --trust-remote-code --tensor-parallel-size 4 --dtype bfloat16 --max-model-len 4096 --enforce-eager

Expected Behavior

Fix the bug and release new version.

Screenshots and logs

No response

Additional Information

No response

[jaredlang]

Has this bug been fixed? I happened to me too.

I found this post.
https://stackoverflow.com/questions/78649446/autogen-groupchat-error-code-openai-badrequesterror-error-code-400

I removed the space in the agent names, and solved the problem.