hmm... understood. means i would send the schema to the llm, wait for a query, execute the query in typesense, and then ask the llm again for a proper response to the user?