Purpose/Significance To explore the method of constructing a Xin’an physician question-answering system based on knowledge graph and large language model(LLM)，so as to promote the systematic inheritance and development of Xin’an medicine. Method/Process Based on data from 1 028 Xin’an physicians，a knowledge graph is constructed. A collaborative model of“large lan. guage model + external knowledge base”is adopted，and a two-layer prompt template is designed. The Cypher-generated prompt tem.plate is responsible for understanding the question and converting it into Cypher statements，while the knowledge integration prompt tem. plate is responsible for integrating structured knowledge and generating the answer. Result/Conclusion Compared with question-answering methods based on LLM，the system built based on this method significantly improves the correctness，accuracy，and ROUGE-L index of the answers，effectively alleviating problems such as“irrelevant answers”and“illusions”.