Abstract: The intelligent management of hospital logistics has long been plagued by core challenges, including data silos, low cross-campus collaboration efficiency, and passive operation and maintenance (O&M) paradigms. Taking Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine as the engineering background, this study proposes and constructs a Logistics Data Repository (LDR) enhancement system integrated with a large language model (LLM). Built upon a robust data foundation that has incorporated nearly 10, 000 IoT sensor endpoints across two campuses, the system innovatively introduces an intelligent cognition and interaction layer powered by a domestically developed LLM, organically bridging foundational perception, data fusion, LLM-driven intelligence, and business applications. Key technologies have been developed, including multimodal O&M data understanding and governance, a natural language-based intelligent Q&A assistant, associative reasoning and early warning for complex scenarios, and digital twin conversational interaction. In practice across the Yangpu and Fengxian campuses of Xinhua Hospital, the system empowers multiple representative scenarios: dual-campus intelligent control, spatial asset management, energy consumption management with intelligent regulation, automatic report generation and intelligent decision-making based on management indicators. These capabilities collectively enable a paradigm shift from scheduled maintenance to predictive maintenance. Application results demonstrate that the system significantly enhances the scientific rigor of management decisions, the proactivity of O&M responses, and user satisfaction, thereby driving a fundamental transition of logistics management from reactive repair to proactive, data-driven decision-making. This study provides a replicable technical framework and practical reference for large public hospitals pursuing data- and intelligence-driven smart logistics.