How to handle the non relevant questions based on context of the RAG application

So I wanted to ask like suppose we have trained the chatbot on a particular set of documents , then I ask some query to that chatbot and it returns me with the most relevant text chunk from the vector db , but suppose after that I ask the chatbot some non relevant but relevant to the conversation question like “can you explain your response in much better way?” , then the chatbot does not gives response. How do we handle such questions ?