[Feat]: Removing TaskUpdater complete/failed/reject helper functions, or add more functions for the rest of the TaskStates

[njbrake]

Is your feature request related to a problem? Please describe.

The API for the TaskUpdater is slightly confusing to me when getting started, because I was seeing some examples making use of the TaskUpdater.complete(message=...) type syntax, while others relied wholly on the TaskUpdater.update_status syntax. I was trying to leverage the TaskState.input_required state, and there was no helper function for that one.

Would you be open to either removing the small helper functions like complete/failed/reject that are only small wrappers around the update_status function? Or, would you be open to adding a few more wrappers to help handling states like input_required?

Describe the solution you'd like

My personal preference would be to remove the helper functions like TaskUpdater.complete Because it seems cleaner to use the update_status function. However, I can understand that maybe having functions like complete may make it easier to read code in some places. If the team wants to keep the helper functions, maybe we could add the missing status helper functions and possibly convert update_status to a private member like _update_status?

Describe alternatives you've considered

There are currently multiple ways to do the same thing, and since both solutions are just a small code difference, aka

updater.update_status(
                TaskState.completed,
                message=new_agent_parts_message(
                    [Part(root=TextPart(text=agent_trace.final_output.result))],
                    task.contextId,
                    task.id,
                ),
                final=True,
            )

vs

updater.complete(
                message=new_agent_parts_message(
                    [Part(root=TextPart(text=agent_trace.final_output.result))],
                    task.contextId,
                    task.id,
                )
            )

It's not clear to me what the added value of the .complete() function is.

Additional context

Found while working on mozilla-ai/any-agent#410

If someone from the google team can provide a 👍 or 👎 on my proposal, I'd be happy to help with the implementation if we agree about taking an action.

Code of Conduct

• I agree to follow this project's Code of Conduct

[holtskinner]

@njbrake Go ahead and make a PR to add more wrappers/helper functions for further functions.
We would like to keep the helper functions available.

[njbrake]

@holtskinner sounds good 🎉 . Does your team have any tips or clarification about what the final means?

My current understanding is that if the A2A Server is not in streaming mode, you have to set final when you send your task update, otherwise it throws a asyncio.queues.QueueEmpty: Queue is closed. error message. In streaming mode, you can choose when to set final=True.

The reason I ask is because for a state like 'input-required', maybe in streaming mode we want final=False, but outside of streaming mode I have to set final=True in order to pass that input-required state back to the client so that I can send a new message. That means, when implementing a helper function for these states I might need to still have a parameter for final, unless I misunderstand something.

At the very least, I can certainly add a helper function for the canceled state, since that should unambiguously be set as final=True.

You can check out the multi-turn agent_executor implementation I'm working on here, if this helps give context at all: mozilla-ai/any-agent#494. It is currently returning a Task object to the client, and is not in streaming mode.