A framework making it effortless to convert any llm model into a reasoning agent like o1 or DeepSeek's r1