I used custom tool to check calendar availability but LLm process first response and ignores the upcoming ones{ "toolCallId": "call_V1iTnnH1jvLsMek6V2EfmfP3", "status": "busy", "success": false, "content": "Unfortunately, that time is already booked. Please suggest another time.", "next_available": "2025-07-14T09:00:00-04:00" } and next slot is free { "toolCallId": "call_wfOLYctQdBGeOKkMYcihZwmF", "status": "free", "content": "Great! The slot is available. I’ll proceed to reserve it for you.'", "success": true } It doesn't update the memory

Set responseDelaySeconds and startSpeakingPlan.waitSeconds to around 4–5 seconds in your assistant's config..