Reinstating complete, which is in a terminal state.

[lukehinds]

When a Task completed, its marked as terminal.

a2a-python/src/a2a/server/request_handlers/default_request_handler.py

Lines 223 to 229 in 4186731

	if task:
	if task.status.state in TERMINAL_TASK_STATES:
	raise ServerError(
	error=InvalidParamsError(
	message=f'Task {task.id} is in terminal state: {task.status.state.value}'
	)
	)

But a Task may appear complete, but could be revived at a later point.

For example, imagine an agent that has calendar scheduling as as skill and presents within a chat client:

"Book me a meeting with [email protected] next week at 14:00"

Agent books meeting , confirms with user, user replied 'great!' task.complete

Four hours later, the user states 'darn, I forgot, I am busy at 14:00, can you move to 16:00'

This will be refused though , as from what I can tell in the A2A framework, once a task reaches a terminal state (like completed, failed, canceled, etc.), it cannot be switched back to a non-terminal state like working. The framework treats terminal states as immutable , I guess for consistency and safety.

At the moment, mine plan is to leave them in working , with a configurable period to clear cache of working -> complete.

[pstephengoogle]

The protocol defines this as a new task not a re-activation of an old task. When sending the follow up message, you can refer to referenceTaskIds to inform the agent this is directly related to a prior task. (https://github.com/a2aproject/A2A/blob/3b891caac0a4897feaa88efab1d31caa1759d33b/types/src/types.ts#L733)