No we don't duplicate, it's just that either your first message is used or LLM model generates the response how it is done normally.