Skip to main content

Data Center Operations AI Tools

Data Center Operations uses AI-assisted workflows through product APIs, Advisor-backed diagnosis, Predictive Maintenance handoff, and AI Engine health or remaining-life calls. The goal is to help operators review risk, prepare maintenance actions, and keep evidence traceable.

Use this page when an implementation team needs to understand which DCOps surfaces can be called by the application, which outputs require review, and how to combine DCOps with DFS, Inspector, and Predictive Maintenance.

Tool layers

LayerPrimary usersAccess boundaryOutput type
/api/v1/dcops/* product APIsData center operations UI and approved backend workflowsdcops.view, dcops.predict.view, dcops.diagnosis.run, dcops.workorder.link, and dcops.feedback.writeDashboards, asset health, predictions, diagnosis, dispatch, BMS mapping, planning, and operations snapshots
Advisor diagnosisDiagnosis workflow and review assistantsProduct service authentication and diagnosis permissionDiagnosis summary, confidence, evidence source, and trace ID
Predictive MaintenanceReliability and maintenance ownerspdm:read, pdm:write, and related product permissionsEquipment health, anomaly review, remaining-life estimates, advisories, and work-order feedback
AI EngineProduct services and controlled backend jobsService authentication and project configurationHealth score, remaining-life estimation, diagnosis support, and engine-mode metadata
DFSData owners and integration teamsDFS permissions for connectors, mappings, datasets, and quality reviewPrepared BMS, alert, meter, equipment, work-order, and signal-history datasets

When an external Agent needs DCOps context, use approved backend workflows and shared /mcp/base/ context for assets, documents, work orders, and evidence. Runtime tool availability should be discovered in the target environment before exposing a workflow.

Data preparation

AI-assisted Data Center Operations depends on a source package that can be reviewed:

DataWhy it matters
Equipment identityConnects dashboards, BMS points, predictions, alerts, work orders, and feedback to the same operating object.
BMS mappingShows which source points drive target fields and whether mapping coverage is acceptable.
Alert historyProvides severity, timing, recurrence, and affected equipment context for diagnosis.
Work-order historyProvides recent actions, open backlog, SLA state, and feedback for closed-loop review.
Health and prediction snapshotsProvides score, risk level, probability, remaining-life range, model version, and engine mode.
Meter and energy readingsSupports PUE/WUE and energy review when definitions, units, and time windows are validated.

Use DFS mappings, Data Quality in DFS Lite, and Prepare Predictive Maintenance Signal History before exposing automated review.

Health and prediction endpoints

Use these endpoints when an asset needs equipment-level health or risk review:

EndpointPurposeReview checks
POST /api/v1/dcops/assets/{assetId}/healthCalculate and persist a health snapshot for one asset.Confirm asset identity, recent alert count, work-order context, source freshness, and engine mode.
GET /api/v1/dcops/assets/{assetId}/health/trendRead health-score trend for a selected time window.Check time range, missing history, and risk-level changes.
GET /api/v1/dcops/health/summaryReview aggregate health distribution across equipment.Confirm site scope and which assets have current snapshots.
POST /api/v1/dcops/assets/{assetId}/predictionsGenerate failure probability and remaining-life output.Review model version, probability windows, remaining-life range, and source assumptions.
GET /api/v1/dcops/assets/{assetId}/predictions/historyRead prediction history for one asset.Compare recent predictions with work orders and field findings.
GET /api/v1/dcops/predictions/high-riskList equipment above the selected risk threshold.Validate threshold, time window, and review owner before dispatch.

Use Predictive Maintenance docs when the workflow requires richer model management, anomaly review, equipment templates, or advisory inbox behavior.

Diagnosis and closed-loop endpoints

Diagnosis endpoints help teams move from alert context to reviewed action:

EndpointPurposeReview checks
POST /api/v1/dcops/diagnosis/from-alert/{alertId}Create a diagnosis record for an alert.Confirm the alert has linked equipment and source timestamps.
GET /api/v1/dcops/alerts/{alertId}/diagnosisRead diagnosis history for an alert.Review diagnosis text, confidence, trace ID, created time, and source.
GET /api/v1/dcops/alerts/{alertId}/closed-loopRead alert, diagnosis, work-order, and feedback evidence together.Confirm diagnosis, action, and feedback are all present before closing review.
GET /api/v1/dcops/work-orders/{workOrderId}/closed-loopRead closed-loop evidence for a work order.Check diagnosis linkage and feedback state.
POST /api/v1/dcops/work-orders/{workOrderId}/feedbackRecord actual root cause, action taken, resolved state, and score.Keep feedback tied to the responsible reviewer and work-order closeout.

Diagnosis output should be treated as a reviewed operating record. The recommended action becomes operational only after the responsible owner approves it.

Planning and dispatch endpoints

Use planning and dispatch endpoints to convert reviewed risk into scheduled work:

EndpointPurposeReview checks
GET /api/v1/dcops/planning/recommendationsRead maintenance candidates for a selected window.Confirm risk score, estimated duration, priority, and asset scope.
POST /api/v1/dcops/planning/generateGenerate a draft plan.Review generated items and conflict count before assignment.
GET /api/v1/dcops/planning/{planId}/conflictsInspect plan conflicts.Check resource or time-window conflicts with shift owners.
POST /api/v1/dcops/planning/{planId}/resolveResolve planning conflicts.Record strategy and remaining conflicts.
GET /api/v1/dcops/dispatch/pendingList open alerts with no open work order.Confirm diagnosis state and action owner.
POST /api/v1/dcops/dispatch/batch-approveApprove selected dispatch candidates.Use after selected alerts and diagnosis evidence have been reviewed.

Operations analytics

Operations analytics endpoints support shift review and handover:

EndpointPurpose
/dashboard/operationsCombined operational queue and throughput dashboard data.
/dashboard/operations/kpi-deltaRecent KPI movement versus the previous period.
/dashboard/operations/queue-agingOpen queue aging buckets.
/dashboard/operations/predictive-queuePredictive risk buckets for 7, 30, and 90 day windows.
/dashboard/operations/sla-ratesResponse and completion SLA rate metrics.
/dashboard/operations/sla-breachesOpen breached work orders sorted for action review.
/dashboard/operations/diagnosis-reuseDiagnosis reuse rate for dispatches.
/dashboard/operations/dispatch-funnelConversion funnel from alert to closed-loop outcome.
/dashboard/operations/snapshotExportable operations snapshot for dashboard and handover.
/dashboard/operations/snapshot.csvCSV export for spreadsheet review.
/dashboard/operations/eventsServer-sent event stream for alerts, work orders, and risk snapshot changes.

Use snapshot exports for operating meetings, shift handover, and post-incident review. Keep the export with the review notes that explain scope and assumptions.

BMS mapping and model operations

EndpointPurposeReview checks
GET /api/v1/dcops/integrations/bms/mappingsRead latest published or draft BMS mapping metadata.Check source, version, status, and updated time.
POST /api/v1/dcops/integrations/bms/mappings/validateValidate mapping rule shape before publish.Check source point, target field, coverage, errors, and warnings.
POST /api/v1/dcops/integrations/bms/mappings/publishPublish a mapping version and create audit evidence.Publish after source owner approval.
GET /api/v1/dcops/model-ops/statusRead active engine mode and status metadata.Record engine mode, degraded state, latency, and update time.
POST /api/v1/dcops/model-ops/engine-modeSwitch standard or NVIDIA mode when authorized.Use project-approved mode, audit the change, and confirm fallback behavior.

Model-ops status is an operational control surface. Treat mode changes as governed configuration changes.

Suggested Agent answer structure

When an AI assistant summarizes DCOps risk, use a stable format:

SectionContent
ScopeTenant, site, data center, data hall, time window, and selected assets.
Current stateDashboard KPIs, open alerts, queue aging, SLA risk, health distribution, and predictive queue.
EvidenceSource records, health snapshots, prediction history, diagnosis records, BMS mapping version, work orders, and feedback.
Recommended reviewRanked actions for the responsible owner, with evidence links and missing data notes.
HandoffWork-order draft, escalation note, planning draft, or request for field verification.

Validation checklist

  • The workflow runs under the correct tenant and DCOps module scope.
  • The caller has the required dcops.* permission for the read, diagnosis, work-order, feedback, or configuration action.
  • Asset identity, BMS points, alerts, work orders, and meter readings resolve to the same site context.
  • AI-assisted output includes source, confidence, trace ID, model or engine mode when available, and review owner.
  • Work-order actions remain tied to approved dispatch or planning evidence.
  • Feedback is recorded after field execution so future diagnosis and planning can learn from the outcome.