Can you confirm if you're using Langchain to query the knowledge base and also which gpt model is behind it. That will maybe give us a better understanding of how to query the kb