After a tool call is triggered, Vapi seem to do another request towards the knowledge base. That seems like a bug. I can see in the logs the following that happens when I ask about a specific in-room dining product, e.g. I would like to order a cheeseburger: 1. Request towards knowledge base to retrieve product info. (trieve) 2. Call to LLM that determines a tool call needs to trigger to fetch modifiers belonging to the requested product. 3. Another request towards knowledge base. (seems overkill) 4. Call to LLM with results of the tool call together with knowledge base results. 5. Response back to user.