| Incident management under the Information Technology Infrastructure Library (ITIL) framework requires structured knowledge representation, rigorous intervention planning, and rapid resolution mechanisms. In this paper, we propose a scientific architecture for building a knowledge base dedicated to a Retrieval-Augmented Generation (RAG) system aligned with ITIL incident management practices. We formalize the knowledge base structure using graph-theoretic and vector space representations, integrate constraint-based modeling for intervention planning, and introduce mathematical formulations governing retrieval, ranking, compliance validation, and controlled response generation. We validate our approach through a comprehensive experimental evaluation, including ablation studies, comparisons with multiple baselines (dense retrieval, hybrid BM25+dense, and prompt-based constraint injection), and disaggregated analysis across varying operational environments (Banking vs. University IT). Results demonstrate that our Constraint Satisfaction Problem (CSP) injection mechanism improves regulatory compliance by 27% in strict environments compared to standard RAG baselines, while maintaining sub-2-second latency at the 95th percentile. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.