Could be achieved with RAG. Like a LLM users instruct which filters to create by a chat.