Abstract:Objective To design a digital human-driven conversational agent system for nutrition education in gastric cancer patients, with the aim of providing professional nutritional consultation and guidance to patients. Methods Based on the conceptual framework of the "corpus cube", a gastric cancer nutrition management corpus was constructed from three dimensions:source, business domain, and modality.A dynamically updating mechanism of the corpus was also established.On this basis, retrieval-augmented ge-neration technology was employed to build a conversational agent system integrated with a digital human interactive interface. Results After three months of system operation, the accuracy of question-answer content reached 90.4%, and the accuracy of question-answer matching scored (4.66±0.50) stars (out of a maximum of 5 stars). Conclusion The effective integration of the gastric cancer nutrition management corpus with digital human conversational agent technology preliminarily demonstrates the feasibility of leveraging digital technology to empower clinical nutrition management, offering professional and convenient nutritional consultation services for gastric cancer patients.