OpenAI runs a Responses backend; the application still executes its own function tools.
Automatic acknowledgement delay in milliseconds.
Responses model, prompt and tools.