Understanding the logic behind LLM responses in REG with LangChain or Llama Index”