Retrieval Augmented Generation models have demonstrated significant potential in enhancing the performance and reducing hallucinations in Large Language Models by incorporating external knowledge sources. However, effectively utilizing both structured and unstructured data in Retrieval-Augmented Generation remains a challenge, as most existing methods retrieve only short, contiguous chunks from a retrieval corpus, limiting the holistic understanding of the overall document context and structure. This paper presents a novel approach to constructing a knowledge graph that seamlessly integrates structured and unstructured data to enhance the performance of Retrieval-Augmented Generation models. Our method involves first converting structured data into a graph model and then merging it with the document context found in unstructured text. This combined data is processed to build a Unified Knowledge graph, providing a rich, interconnected information source for the Retrieval-Augmented Generation model. To evaluate our approach and proposed model, we conducted extensive experiments on a variety of tasks, including question-answering and complex reasoning. The results demonstrate that our Unified knowledge Graph RAG model significantly outperforms traditional RAG models, achieving higher accuracy and better handling of intricate queries. Notably, on benchmarks such as QuALITY, our model shows substantial improvements in performance metrics. This research underscores the potential of integrating structured and unstructured data into a unified knowledge graph to advance the capabilities of RAG models, offering a promising direction for future developments in Artificial Intelligence and Natural Language Processing.