Toward a Context-Aware AIOps Framework for High-Energy Physics Computing Clusters: A Design Proposal with Isolation Forest Scoring and Operator-Controlled Response

High-energy physics (HEP) computing facilities combine heterogeneous compute nodes, distributed storage subsystems, multi-domain networks, job schedulers and loosely coupled services. Field-validated incident labels are scarce in practice, and legitimate shifts in the workload mix can also raise anomaly indicators. This paper puts forward a context-aware AIOps architecture; Isolation Forest ranks anomalous telemetry windows, full data lineage is kept, and human site operators hold the authority over any response. The stack has six functional tiers, that is, telemetry ingestion, data-quality governance, a temporally valid context store, multi-modal hybrid analytics, event triage and a policy-enforced response executor. Single alert triggers are merged into incident episodes, and the scores only set the review priority; they do not estimate how likely a failure is. Four deployment gates, from offline retrospective analysis to approval-gated actions, limit how far the system may act. The paper reports a design and an evaluation plan; no new operational data are analysed, and no performance gain in production is claimed.