Question: Can agents create steps or tasks for itself?
thomafred opened this issue · comments
Does the protocol any restrictions to whether an agent can create tasks or steps for itself? How does this relate to the is_last-flag?
The analogy is that a task is a ChatGPT chat window, and and a step is a message to the chatGPT (in this case the agent).
Another agent can definitely create a task and steps onto your existing agent.
When the existing agent identifies that the step is the last_step, it will return to the client with last_step=True
I am closing the issue, please reopen if you have additional questions!
@hackgoofer Thanks for the clarification.
I guess my confusion revolves around the last_step-flag. In what circumstances does the agent set this flag to true, and what implications does this have? Can we for example start cleaning up resources at this point?
In half-duplex systems such as walkie-talkies for example, protocol dictates that the person starting a conversation is also the one ending it.