The Core Loop: Simulate, Observe, Monitor, Optimize
Tuner is built around a continuous improvement loop:Simulate
Stress-test your agent with Call Simulation before launch and after every change. The same Evals that score production calls score every simulated call.
Key Concepts at a Glance
| Concept | Description | Learn More |
|---|---|---|
| Evaluations & Guardrails | Evaluations analyze a call against a set of criteria to determine its quality. Guardrails are rules that define what “bad” looks like—when breached, they trigger red flags or alerts. | Introduction to Evaluations & Guardrails |
| Metrics | Quantitative measurements for every call, like talk time, silence duration, and sentiment. | Introduction to Pre-defined Metrics |
| Red Flags | Tags triggered when a critical issue is detected, like a hallucination or broken flow. | Best Practices for Agent Monitoring |
| Alerts | Proactive notifications sent when a specific event occurs, like a spike in red flags. | Introduction to Real-Time Alerts |
| Call Simulation | AI-powered test calls with your agent, dialing inbound agents over SIP or triggering your outbound call flow, scored against your production Evals. Use before launch, after changes, and to validate fixes. | Introduction to Call Simulation |
Next Steps
Dive deeper into the terminology used throughout the Tuner platform.Glossary of Key Terms
A comprehensive glossary defining all key terms used in Tuner.