This system applies large language model technology, choosing ChatGLM2 as the base model, and obtains 18,905 pieces of information covering the domains of admissions and table understanding through web information retrieval, forming an admissions domain dataset. The system fine-tunes the large language model ChatGLM2 using the admissions domain dataset to optimize its performance in natural language understanding and generation. During the fine-tuning process, the system focuses on optimizing the large language model for table understanding and admissions information Q&A functionalities. The system is implemented using an open architectural framework and modular design principles, incorporating the LangChain framework and connecting the word embedding processing node to the Ernie3.0 model. The open system architecture ensures that the admissions system can flexibly meet diverse admissions consulting needs and allows for continuous updates and optimization upgrades. The system service interface uses Gradio, which significantly simplifies the demonstration and sharing process of machine learning models. By providing a simple API, Gradio enables developers to quickly create interactive web interfaces. The innovation of this system lies in introducing large language models to the field of admissions consulting, leveraging their powerful language understanding and generation capabilities to provide users with smarter, customized services. This contributes new ideas and methods to the future development of the admissions consulting field, pushing admissions consulting services towards more intelligent and efficient directions.