This research presents the design, implementation, and evaluation of a hybrid semantic question answering (QA) system developed specifically for the historical knowledge domain of the Sumedang Larang Kingdom using Bahasa Indonesia. The system integrates three complementary layers of knowledge representation retrieval-augmented generation (RAG), knowledge graphs (KG), and ontology-based reasoning within a modular architecture, guided by a large language model (LLM) that performs both question interpretation and answer synthesis. A total of 100 questions covering seven distinct types (what, who, when, where, list, numeric, and yes/no) were used to evaluate the system’s performance under four di!erent configurations: LLM+RAG, LLM+KG, LLM+Ontology, and a Hybrid setup. The results showed that each module contributed unique strengths, and the Hybrid configuration achieved the highest overall accuracy of 83.77%. The findings validate the hypothesis that multi-source semantic integration improves both the accuracy and contextual depth of answers in low-resource and culturally specific settings. This study contributes to the field of semantic QA by o!ering a practical and extensible framework that combines unstructured retrieval, structured data, and formal reasoning for intelligent question answering in Bahasa Indonesia.