[BUG] A2A - completed requested ignored

[chk0301]

Description

hi, A2A remote agent call responds message properly with status "completed", but A2A server is ignoring it and delegating same request to the same agent again. Remote agent respond again with same answer and status "completed. This repeats until multi-turn = 10 and Server agent terminates (max - turn reached)

Steps to Reproduce

use existing code A2A server crew and A2A remote crew

Expected behavior

after response with status "completed" is received, the server agent does not delegates again,

Screenshots/Code snippets

here is where the "loop" get started

A2A/wrapper.py _delegate_to_a2a(

        if a2a_result["status"] in ["completed", "input_required"]:
            final_result, next_request = _handle_agent_response_and_continue(
                self=self,
                a2a_result=a2a_result,
                agent_id=agent_id,
                agent_cards=agent_cards,
                a2a_agents=a2a_agents,
                original_task_description=original_task_description,
                conversation_history=conversation_history,
                turn_num=turn_num,
                max_turns=max_turns,
                task=task,
                original_fn=original_fn,
                context=context,
                tools=tools,
                agent_response_model=agent_response_model,
            )

            if final_result is not None:
                return final_result

            if next_request is not None:
                current_request = next_request

            continue

Operating System

Windows 11

Python Version

3.12

crewAI Version

1.3.0

crewAI Tools Version

n/a

Virtual Environment

Venv

Evidence

code attached

Possible Solution

why when status = completed, the process of

final_result, next_request = _handle_agent_response_and_continue(

needs to be done again .. it's final.

Additional context

server agent (standalone, just execute. ensure .env contains model, key and url. I was using gpt.4.1

MultiAgentChat_light.py

and remote agent. (ensure .env exists), start main.py

agent.py
agent_executor.py
main.py

[greysonlalonde]

This indicates that your agent does not respect the status provided by the remote agent, which is by design. If a task is not completed in the local agents perspective, it will try again. Does the local agent appear to behave that way? Can test a little later

[chk0301]

hi @greysonlalonde, yes - but it's nothing I can do on the agent to change this behavior. I followed the instructions provided in https://docs.crewai.com/en/learn/a2a-agent-delegation
It's similar to the basic configuration.

The remote agent works with googles standard implementation. So seam to be something within the crewai logic, when remote agents response is received. as per my basic debugging, I came to

_handle_agent_response_and_continue() which is called from A2A/wrapper.py _delegate_to_a2a() under the IF condition if a2a_result["status"] in ["completed", "input_required"]:

thank you for your help !

[greysonlalonde]

hi @greysonlalonde, yes - but it's nothing I can do on the agent to change this behavior. I followed the instructions provided in https://docs.crewai.com/en/learn/a2a-agent-delegation It's similar to the basic configuration.

The remote agent works with googles standard implementation. So seam to be something within the crewai logic, when remote agents response is received. as per my basic debugging, I came to

_handle_agent_response_and_continue() which is called from A2A/wrapper.py _delegate_to_a2a() under the IF condition if a2a_result["status"] in ["completed", "input_required"]:

thank you for your help !

We can surface a flag here to respect the response status or not, I think that should cover this.

[greysonlalonde]

@chk0301 we've surfaced a flag to enable this behavior, will be on the upcoming release.