Skip to main content
Observability dashboards give you a live view of how your agent performs in production. They aggregate evaluation data from real customer conversations and surface trends, anomalies, and health signals so your team can act on what matters.

What Observability Dashboards Show

Every call evaluated through the observability pipeline feeds into your dashboards. The data includes:
  • Custom Metric scores — tracked over time so you can spot trends and regressions
  • Hallucination and redundancy rates — built-in quality signals aggregated across all evaluated calls
  • Latency metrics — average agent latency, time to first utterance, and per-turn response times
  • Call volume and status — how many calls are being evaluated and how they’re distributed
  • Alert badges — visual indicators when metrics cross configured thresholds

Common Use Cases

Production Pass-Rate Monitoring

Track your most important Custom Metrics across all production calls. Set up a dashboard that shows daily or weekly pass rates for metrics like compliance adherence, resolution success, or customer satisfaction. Use it in team standups or leadership reviews to confirm your agent is meeting its quality bar.

Customer Sentiment Analysis

Build dashboards around customer-facing metrics — sentiment scores, escalation rates, and interaction quality. These help you understand how customers experience your agent over time and catch shifts before they become problems.

A/B Testing Agent Releases

When rolling out a new prompt version or model change, compare metric scores between the old and new variants. Observability dashboards let you filter by metadata fields like prompt_version or experiment_group to see side-by-side performance.

Operational Health

Monitor operational signals like average latency, token usage, and error rates. These dashboards help engineering teams catch infrastructure-level issues — slow model responses, integration timeouts, or transcript quality degradation — before they affect customer experience.

Structuring Your Dashboards

StrategyWhen to use
One dashboard per agentTeams with a small number of agents that need unified monitoring
One dashboard per business functionOrganizations where different teams own different quality domains (support, sales, compliance)
One dashboard for leadershipHigh-level view of pass rates and volume across all agents
One dashboard for engineeringOperational metrics like latency, errors, and token usage
Pair observability dashboards with alerts so your team gets notified when key metrics cross a threshold, rather than relying on someone checking the dashboard manually.

Next Steps

Alerts

Configure threshold-based notifications for production metrics.

Custom Metrics

Define the evaluation criteria your dashboards track.

API Integration Tutorial

Connect your production pipeline to start feeding dashboards.

Dashboards Overview

Learn about both simulation and observability dashboard surfaces.