Hey everyone! I've been building voice agents for a while and always struggled with the "black box" problem—not knowing exactly where the latency was coming from or why a specific turn failed. So I built a Voice Agent Observability Platform to fix this. It integrates directly with Vapi. What it does (Beta): Unified Dashboard: See all calls from Vapi in one place. Latency Breakdown: Estimates ASR, LLM, and TTS latency per turn (inferred from available data). Cost Tracking: Real-time spend monitoring. Note: Some metrics are currently inferred as we refine the data models. I'm releasing it in Public Beta to get feedback. It's still early days, so expect some rough edges! https://hevo.dev/