Automated storytelling in cultural-heritage contexts requires more than digitized collections and static metadata: it demands formal models of historical entities, events and social relations, as well as computational mechanisms capable of transforming structured knowledge into coherent narratives. This paper presents HistoricO-Composer, a framework that uses an ontology as a structured foundation for generating short, context-aware narratives from archival collections. The approach links three components: semantically annotated items, query patterns that retrieve relevant material, and narrative templates that turn the selected information into concise stories. A constraint layer ensures that dates, roles and provenance remain consistent, so that the output reflects the historical record rather than speculation. A working prototype shows that this method can support historians in exploring themes, presenting collections and creating coherent narrative assets across different domains. The goal is to offer cultural institutions and researchers a practical tool to communicate archival knowledge in a transparent and sustainable way.