npm install @reaatech/agent-eval-harness-observability
# or
pnpm add @reaatech/agent-eval-harness-observability

import {
  getLogger,
  getTracingManager,
  getMetricsManager,
  getDashboardManager,
} from "@reaatech/agent-eval-harness-observability";
 
// Structured logging with automatic PII redaction
const logger = getLogger();
logger.info({ runId: "eval-123", trajectories: 50 }, "Evaluation started");
logger.error({ err: new Error("Connection lost") }, "Judge API call failed");
 
// Metrics recording
const metrics = getMetricsManager();
metrics.recordRun("success", 1);
metrics.recordTrajectories("production", 50);
metrics.recordJudgeCall("claude-opus", "success");
metrics.recordJudgeCost("claude-opus", 0.0234);
metrics.recordGateResult("overall-quality", true);
metrics.recordLatencyP99("evaluation", 3200);
 
// Dashboard with trend analysis and alerting
const dashboard = getDashboardManager();
dashboard.recordRun({
  overallMetrics: { overallScore: 0.87, avgCostPerTask: 0.05, latencyP99: 3200 },
  summary: { totalTrajectories: 50, passRate: 92 },
  metricBreakdown: { faithfulness: { avgScore: 0.85 } },
});
 
console.log(`Quality trend: ${dashboard.getSummary().trends.score}`);
console.log(`Active alerts: ${dashboard.getAlerts().length}`);

import { getLogger, createChildLogger, setGlobalRunId, getGlobalRunId } from "@reaatech/agent-eval-harness-observability";

import { getMetricsManager, recordMetric, incrementCounter } from "@reaatech/agent-eval-harness-observability";

import { getTracingManager, withTracing, addSpanAttributes } from "@reaatech/agent-eval-harness-observability";

import { getDashboardManager } from "@reaatech/agent-eval-harness-observability";

interface DashboardSummary {
  totalRuns: number;
  currentScore: number | null;
  currentPassRate: number | null;
  currentCostPerTask: number | null;
  currentLatencyP99: number | null;
  activeAlerts: number;
  trends: {
    score: "up" | "down" | "stable";
    passRate: "up" | "down" | "stable";
  };
}

import { withTracing, addSpanAttributes } from "@reaatech/agent-eval-harness-observability";
 
const result = await withTracing(
  "custom_validation",
  async (span) => {
    // Span is active throughout this block
    addSpanAttributes({ validation_type: "schema", schema_version: "2.1" });
 
    const isValid = await validateInput(payload);
    return { isValid, timestamp: Date.now() };
  },
  { "custom.attribute": "value" },
);
 
// Span automatically ends — success status on return, error on throw

import { getDashboardManager } from "@reaatech/agent-eval-harness-observability";
 
const dashboard = getDashboardManager({
  alertThresholds: { qualityScore: 0.85, costPerTask: 0.03, latencyP99: 3000, passRate: 0.90 },
});
 
// Record a run
dashboard.recordRun({
  overallMetrics: { overallScore: 0.82, avgCostPerTask: 0.04, latencyP99: 4500 },
  summary: { totalTrajectories: 100, passRate: 88 },
  metricBreakdown: {},
});
 
// Check summary
const summary = dashboard.getSummary();
console.log(`Runs: ${summary.totalRuns}`);
console.log(`Score trend: ${summary.trends.score}`);
console.log(`Active alerts: ${summary.activeAlerts}`);
 
// Inspect alerts
for (const alert of dashboard.getAlerts()) {
  console.log(`[${alert.level}] ${alert.metric}: ${alert.message}`);
}
 
// Generate full dashboard panels (chart, stat, alert)
const panels = dashboard.generateDashboard();
for (const panel of panels) {
  console.log(`${panel.title} (${panel.type}): ${panel.metrics.length} metrics`);
}

import { getLogger, createChildLogger, setGlobalRunId } from "@reaatech/agent-eval-harness-observability";
 
const logger = getLogger();
setGlobalRunId("eval-run-42");
 
// All subsequent log lines include run_id: "eval-run-42"
logger.info({ taskType: "password-reset" }, "Starting evaluation");
 
// Create per-component child loggers
const judgeLogger = createChildLogger({ component: "judge" });
judgeLogger.info({ model: "claude-opus", metric: "faithfulness" }, "Judge evaluating");
 
// Errors carry stack traces
try {
  await doWork();
} catch (err) {
  logger.logError(err as Error, { taskId: "task-7" });
}

import { getMetricsManager } from "@reaatech/agent-eval-harness-observability";
 
const metrics = getMetricsManager();
 
metrics.recordBatchMetrics({
  runs: { status: "success" },
  trajectories: { dataset: "production" },
  judgeCalls: { model: "claude-opus", status: "success" },
  judgeCost: { model: "claude-opus", cost: 0.0234 },
  costPerTask: { taskType: "password-reset", cost: 0.0045 },
  gateResult: { gateName: "overall-quality", passed: true },
  latencyP99: { component: "evaluation", latencyMs: 3200 },
});