Voice Agents — TraceVox

Run scenario-based evaluation campaigns against your voice agent and get findings attributed to their actual cause — target agent, simulator, or infrastructure — with severity and confidence.

• Campaign metrics: scenario success, target-agent bugs, high-severity counts, median/P95 latency, interruption rate.
• Per-run: outcome, quality score, turn-taking, interruptions, peak risk.
• Speaker activity timeline and audio envelope (RMS energy — no raw audio stored).
• Transcript with timestamps aligned to the event timeline; risk trajectory across the run.
• Early Warning Engine: mid-call risk predictions with weighted contributing signals and recommended actions — the same engine that triages the LLM surface.

LLM Applications · Sign in

TraceVox Research home · Public Research Library · Documentation · llms.txt