hello! I am curious - is it possible to use OpenAIAgent or another variation of agent to decouple response from function execution? My end goal is to have "approve" flow before each function execution.
The only option I see now is to copy the agent class and figure out some changes there