Evenfire is an intelligence layer platform for running LLM agents that take real actions, on your infrastructure
TypeScript
/**
* Prometheus metrics for the Workload Recipes Controller (WRC).
*
* Uses a dedicated Registry to avoid polluting the global default registry.
* Controller metrics keep the clerum_wrc_ prefix. Workflow runtime metrics use
* workflow_* names so coordinator and broker dashboards can share labels.
*/
import { Counter, Gauge, Histogram, Registry } from 'prom-client'
const REGISTRY_KEY = Symbol.for('clerum.wrc.prometheus.registry')
type MetricsGlobal = typeof globalThis & { [REGISTRY_KEY]?: Registry }
const metricsGlobal = globalThis as MetricsGlobal
const newRegistry = !metricsGlobal[REGISTRY_KEY]
export const registry = (metricsGlobal[REGISTRY_KEY] ??= new Registry())
if (newRegistry) registry.setDefaultLabels({ service: 'workflow-recipes' })
type MetricOptions = { name: string; help: string; labelNames?: readonly string[] }
function counter(options: MetricOptions): Counter<string> {
const existing = registry.getSingleMetric(options.name)
if (existing) return existing as Counter<string>
return new Counter({
...options,
labelNames: [...(options.labelNames ?? [])],
registers: [registry],
})
}
function gauge(options: MetricOptions): Gauge<string> {
const existing = registry.getSingleMetric(options.name)
if (existing) return existing as Gauge<string>
return new Gauge({
...options,
labelNames: [...(options.labelNames ?? [])],
registers: [registry],
})
}
function histogram(options: MetricOptions & { buckets: number[] }): Histogram<string> {
const existing = registry.getSingleMetric(options.name)
if (existing) return existing as Histogram<string>
return new Histogram({
...options,
labelNames: [...(options.labelNames ?? [])],
registers: [registry],
})
}
export const recipesTotal = gauge({
name: 'clerum_wrc_recipes_total',
help: 'Total number of WorkflowRecipes by phase',
labelNames: ['phase'] as const,
})
export const reconciliationsTotal = counter({
name: 'clerum_wrc_reconciliations_total',
help: 'Total WorkflowRecipe reconciliations',
labelNames: ['result'] as const,
})
export const mcpSessionsActive = gauge({
name: 'clerum_wrc_mcp_sessions_active',
help: 'Number of active MCP sessions',
})
export const policyViolationsTotal = counter({
name: 'clerum_wrc_policy_violations_total',
help: 'Total policy violations detected',
labelNames: ['policy', 'rule'] as const,
})
export const workflowStepDurationSeconds = histogram({
name: 'workflow_step_duration_seconds',
help: 'Workflow step execution duration by executor type and step kind',
labelNames: ['executor', 'stepKind', 'status'] as const,
buckets: [0.05, 0.1, 0.25, 0.5, 1, 2.5, 5, 10, 30, 60, 120, 300],
})
export const workflowStepTotal = counter({
name: 'workflow_step_total',
help: 'Workflow steps completed by executor type and final status',
labelNames: ['executor', 'stepKind', 'status'] as const,
})
export const governedTraceEnqueuedTotal = counter({
name: 'clerum_wrc_governed_trace_enqueued_total',
help: 'Best-effort governed trace events accepted by the WRC local buffer',
labelNames: ['family', 'type'] as const,
})
export const governedTraceDroppedTotal = counter({
name: 'clerum_wrc_governed_trace_dropped_total',
help: 'Best-effort governed trace events dropped by the WRC local buffer',
labelNames: ['family', 'type', 'reason'] as const,
})
export const governedTraceFlushesTotal = counter({
name: 'clerum_wrc_governed_trace_flushes_total',
help: 'WRC governed trace flush outcomes',
labelNames: ['family', 'result'] as const,
})
export const governedTraceRetriesTotal = counter({
name: 'clerum_wrc_governed_trace_retries_total',
help: 'WRC governed trace retries scheduled after a failed flush',
labelNames: ['family', 'type'] as const,
})
export const governedTraceGapsTotal = counter({
name: 'clerum_wrc_governed_trace_gaps_total',
help: 'WRC governed trace evidence gaps that block complete trace or cost coverage',
labelNames: ['family', 'type', 'reason'] as const,
})