Large language models are changing how recommender systems interact with users, moving beyond predicting the next item a user might like to understanding their needs and adapting through natural language feedback. However, no datasets exist to explore these new capabilities, so a new benchmark dataset called Reviews Enhanced with GEnerative Narratives (REGEN) was developed. REGEN incorporates item recommendations, natural language features, and personalized narratives, allowing for the exploration and benchmarking of new recommender architectures. The dataset was created by augmenting the Amazon Product Reviews dataset with synthetic user critiques and narratives generated using the Gemini 1.5 Flash model. REGEN enables the evaluation of models that incorporate user feedback and output natural language consistent with the recommendations. Experiments show that large language models trained on REGEN can effectively generate both recommendations and contextual narratives, achieving performance comparable to state-of-the-art recommenders and language models. The dataset includes critiques, which allow users to express their preferences, and narratives, which provide rich contextual information about recommended items. Two baseline architectures were developed to explore different modeling approaches: a hybrid system and a fully generative model called LUMEN. The results show that REGEN can meaningfully challenge and differentiate models across both recommendation and generation tasks, and that incorporating user critiques into the input consistently improves recommendation metrics. REGEN provides a fundamental resource for studying the capabilities of conversational recommender models, advancing conversational recommendation by integrating language as a fundamental element.