I have two questions related to it: 1. It always seem to look in Trieve for context before passing the question to the LLM which seems like a waste of latency for questions that doesn't require the knowledge base. I would like only certain questions to be using the knowledge base. Any idea if this is possible? 2. I have a knowledge base with product information but would like to add some additional information to each chunk which shouldn't be used when searching results but should be used when retrieving the chunk. e.g. I have a product breakfast which has modifiers like drinks, bread etc. If I chunk it with these modifiers the search results won't be accurate. It should only search on the product itself, but also returns modifiers for that chunk.