I tried using the {{conversation.SummaryAgent.transcript}} and feeding it to the openAI API bit I think it has a limit on how much context it can store