Retrieval Augmented Generation (RAG) stands as a promising paradigm in Natural Language Processing (NLP), integrating information retrieval and generative modeling to produce contextually relevant and factually accurate responses to user queries. This study explores the application of RAG methodology to the retrieval and synthesis of historical milestone events from Wikipedia’s “On this day” and “Today sections.” Leveraging clustering algorithms such as K-means, we organize the dataset into coherent clusters and employ cosine similarity metrics to extract relevant context events. These events are then utilized to augment the input prompt for Google Gemma 7b, the Large Language Model (LLM) incorporated in the RAG pipeline, facilitating the generation of coherent and informative responses tailored to user queries. Our approach addresses the challenge of preventing hallucinations in LLM outputs and enables the adaptation of models to work with custom data, thereby enhancing their utility across diverse domains. We conducted a comprehensive evaluation using both quantitative metrics (Precision, Recall, BLEU, and ROUGE) and qualitative analysis. Our qualitative analysis of the research focused on three key aspects: the coherence, relevance, and completeness of the model's responses. Potential applications of this work include customer support, email analysis, internal documentation chat, and textbook Q&A.