Querying data and documents using LLM models with Kernel Memory