
100+ voice logs per session handled live
Open-source release
Problem
Voice-first products generate hundreds of log events per session. Founders ended up grep-ing JSON to figure out what users said, what failed, and what worked.
Solution
A live dashboard that streams voice events the moment they happen, with per-request inspection and pipeline-health-at-a-glance views.
Impact
Operators stopped parsing JSON by hand. Investigating a broken voice flow went from minutes of grep to seconds of clicking.
What I built
- Real-time API integration that handles frequent voice-log bursts without dropping events
- Per-request inspector for transcripts, payloads, and timings
- Open-sourced: shipped publicly on GitHub
Stack
ReactTypeScriptTailwind CSSREST APIsReal-time data