Stop gracefully stops the scheduling engine.

String returns a human-readable SLO status.

TriggerFailover manually triggers failover for a component.

UpdateNodes replaces the cached node list.

UpdateQValue updates the Q-table based on observed reward (Q-learning update rule) Q(s,a) ← Q(s,a) + α * [r + γ * max_a'(Q(s',a')) - Q(s,a)]

============================================================================ Adaptive Scheduling Interval ============================================

Run multi-method anomaly detection on a metric.

estimateNodeCost estimates hourly cost based on GPU type and count.

parsePrometheusText parses Prometheus text exposition format

============================================================================ SDK type -> local type conversion =======================================

AddAfter adds a request to the queue after a specified delay.

AddSilence adds a silence rule.

============================================================================ Score Adjustment ========================================================

AdvancePromotion moves the pipeline to the next stage.

BuiltinMigrations returns all CloudAI Fusion schema migrations in version order. These are applied automatically on server startup via Migrator.Up().

Checksum computes SHA-256 of a migration's up SQL.

ClassifyAndRouteAlert classifies an alert and routes it to the appropriate channels.

CompleteBackup marks a backup as completed with size and checksum.

============================================================================ Cross-Cluster Load Balancing ============================================

============================================================================ Schedule Operations =====================================================

DefaultConfig returns default event bus configuration.

DefaultCostOptimizerConfig returns production-ready defaults

DefaultManagerConfig returns sensible defaults.

DefaultPredictiveScalingConfig returns production-ready defaults

DefaultQueueManagerConfig returns production-ready defaults

DeployModel deploys an AI model to an edge node via real communication protocol

============================================================================ ACK API Methods =========================================================

DetectFaults runs all enabled fault detectors and returns detected faults.

Discover fetches the OIDC discovery document for a provider

EnabledCount returns the number of currently enabled flags.

Enqueue adds an item to the queue, maintaining priority order.

EvaluateCanary checks canary health and progresses or rolls back.

ForecastDemand generates a resource demand forecast for a cluster.

Get returns the version information populated at build time.

GetLatest returns the most recent collected metrics

GetStats returns a snapshot of engine scheduling statistics.

HandleAdmissionExtension routes admission-related actions

HasPermission checks if a role has a specific permission

Increment advances the clock for the given node.

Init initializes the OpenTelemetry tracing pipeline. Returns a Provider whose Shutdown() must be called on server stop.

Node Operations

ListSSOConfigs returns all SSO configs for a tenant

ListTickets returns all tickets for a tenant

Matches returns true if the event matches the filter.

Metrics returns current queue statistics.

Name returns the reconciler name.

Name returns the reconciler name.

Name returns the reconciler name.

NewBreakerRegistry creates a registry with shared default config.

NewCostAwareScorePlugin creates the cost scorer.

NewCostOptimizer creates a new cost optimizer

NewMemoryConsumer creates an in-memory message consumer.

NewPredictiveScaler creates a new predictive scaling manager

NewQueueManager creates a new unified queue manager

NewRedisCache creates a Redis-backed cache. Falls back to in-memory if Redis is unavailable.

NewRegistry creates a metrics registry wrapping the Prometheus default registry.

NewSLOTracker creates a new SLO tracker.

Read retrieves a value.

Reconcile performs a single reconciliation pass for a Cluster object. It compares the desired state (from the platform's database/spec) against the ac

RecordEvent allows external callers (reconcilers) to record events.

============================================================================ Log Ingestion Stats =====================================================

RegisterDetector adds a fault detector.

SaveSnapshot persists the current queue and running set to the store.

ScanClusterPods scans all pods in a cluster for security misconfigurations This performs REAL K8s API queries to analyze running pod specs.

============================================================================ Q-Learning Core =========================================================

Stats returns election runtime statistics.

ValidateConfig validates the HA configuration.

ValidateEnum checks that a value is one of the allowed values.

ValidateMaxLength checks that a string does not exceed a maximum length.

deployLocal creates instances tracked in memory

getLimiter returns the rate.Limiter for a given key, creating one if needed.

isInsecureDefault checks if a value matches known placeholder/default secrets. Covers .env.example placeholders, common weak passwords, and CI/CD defa

============================================================================ Conversion Helpers ======================================================

============================================================================ memoryElector — Run / IsLeader / GetLeader / Resign / Stats =============

============================================================================ GKE API Methods =========================================================

parseMetricLine parses a single Prometheus metric line Format: metric_name{label1="val1",label2="val2"} value [timestamp]

recordAudit records an admission audit entry

============================================================================ Internal Helpers ========================================================

============================================================================ Timer Management ========================================================

============================================================================ Utility helpers =========================================================

============================================================================ SDK type -> local type conversion =======================================

ActiveProfile detects the active profile from the CLOUDAI_FEATURE_PROFILE env var. Returns ProfileStandard if not set or invalid.