Documentation

JobMetrics

Table of Contents

Classes

Conversation
End-to-end floor dynamics, measured off the audio timeline — the ground truth of what the caller hears, independent of agent self-reports.
LLM
Generated from protobuf message <code>livekit.SimulationRun.JobMetrics.LLM</code>
Simulator
Was the call spoiled by the simulator, not the agent? Diagnostic only — never folded into the scores; run aggregation excludes flagged calls from the two headline scores.
STT
The agent's perception of the caller.
TTS
The agent's voice.
Turn
Generated from protobuf message <code>livekit.SimulationRun.JobMetrics.Turn</code>
On this page

Search results