Objective Primary healthcare is the cornerstone of the medical and health system， where traditional Chinese medicine （TCM） services play a pivotal role with their advantages of simplicity， accessibility， affordability and efficacy， and the distinctive feature of syndrome differentiation and treatment. However， primary-level healthcare faces structural challenges， including talent shortages， uneven resource allocation， outdated technology， and low service efficiency. The emergence of large language models （LLM） offers new opportunities for intelligent transformation， yet semantic adaptation barriers persist between LLM and TCM''s abstract terminology， classical literature， and complex syndrome differentiation logic， hindering practical deployment.Method This paper proposes the concept of Intelligent Primary Healthcare and explores an LLM-driven innovation pathway for primary TCM services. First，TCM corpus and knowledge graph are constructed to enhance the model''s semantic comprehension. Second， a multimodal intelligent diagnosis and treatment framework integrating the four diagnostic methods （inspection， listening and smelling， inquiry， and palpation） is designed. Guided by the human–machine collaborative governance philosophy， an application model covering chronic disease intervention in villages， intelligent operation of TCM clinics， and the role of primary health gatekeepers is established. Furthermore， systematic strategies are proposed to address challenges such as data scarcity， insufficient computing power， and ethical risks， including lightweight deployment， edge computing， and ethical norms.Result The results clarify the applicability of LLM in three primary scenarios and construct a multimodal diagnosis and treatment architecture embedded with TCM syndrome differentiation thinking， overcoming the limitations of conventional artificial intelligence in multi-source information integration. A lightweight deployment scheme is developed to adapt to primary computing constraints. Additionally，the preliminary TCM-specific corpus system and the ethical framework are formed to ensure safe implementation.Conclusion LLM provide a feasible path for the intelligent reconstruction of primary TCM services and exhibit remarkable potential in building a "Human-Machine Co-governance" primary TCM ecosystem. This model improves the accessibility and standardization of primary TCM services， and provides theoretical and practical support for the modernization of TCM and the advancement of tiered healthcare. It thus holds broad promotion value in relevant fields.