OPENING PARAGRAPH:
We built Quiet Hours Valet's command center in 72 hours—a production voice AI system handling 8 concurrent agents, 8 automation workflows, and real-time dispatch across our valet trash operation. The stack: Supabase Edge Functions as our VAPI webhook processor, a normalized 8-table Postgres schema for agent state management, 5 Next.js frontend pages for operator control, and Vercel for zero-latency deployment. This article breaks down the architecture decisions, the Postgres schema design that scaled from prototype to production, and how we structured Edge Functions to handle async voice callbacks without timeout limits—lessons learned building AI infrastructure for a logistics-first business.
Top comments (0)