This paper proposes a novel framework that integrates large language models (LLMs) with graph neural networks (GNNs) to enhance recommendation systems by leveraging edge information in graph data. The framework aims to improve the accuracy and personalization of recommendations by utilizing the contextual understanding capabilities of LLMs and the relationship extraction functions of GNNs. The key innovation is the design of a new prompt construction framework that integrates relational information of graph data into natural language expressions, enabling LLMs to more intuitively grasp the connectivity information within graph data. Additionally, the framework introduces graph relationship understanding and analysis functions into LLMs to enhance their focus on connectivity information in graph data. By enhancing the understanding of graph relationships, the framework provides more comprehensive, accurate, and personalized recommendations. The proposed method is evaluated on real-world datasets, demonstrating its ability to understand connectivity information in graph data and improve the relevance and quality of recommendation results. The framework is implemented using GPT-2 as the base model, with pre-training on crowd contextual prompts and fine-tuning with personalized predictive prompts. The results show that the proposed method outperforms existing recommendation systems in terms of recommendation accuracy and personalization. The framework is designed to be integrated into existing recommendation systems, providing a new technological path for developing more efficient and intelligent recommendation systems. The study also highlights the importance of incorporating graph structure information into LLMs for improving recommendation performance.This paper proposes a novel framework that integrates large language models (LLMs) with graph neural networks (GNNs) to enhance recommendation systems by leveraging edge information in graph data. The framework aims to improve the accuracy and personalization of recommendations by utilizing the contextual understanding capabilities of LLMs and the relationship extraction functions of GNNs. The key innovation is the design of a new prompt construction framework that integrates relational information of graph data into natural language expressions, enabling LLMs to more intuitively grasp the connectivity information within graph data. Additionally, the framework introduces graph relationship understanding and analysis functions into LLMs to enhance their focus on connectivity information in graph data. By enhancing the understanding of graph relationships, the framework provides more comprehensive, accurate, and personalized recommendations. The proposed method is evaluated on real-world datasets, demonstrating its ability to understand connectivity information in graph data and improve the relevance and quality of recommendation results. The framework is implemented using GPT-2 as the base model, with pre-training on crowd contextual prompts and fine-tuning with personalized predictive prompts. The results show that the proposed method outperforms existing recommendation systems in terms of recommendation accuracy and personalization. The framework is designed to be integrated into existing recommendation systems, providing a new technological path for developing more efficient and intelligent recommendation systems. The study also highlights the importance of incorporating graph structure information into LLMs for improving recommendation performance.