OPENING PARAGRAPH:

We just shipped a production voice command center handling real-time operations across a vertically integrated business—5 frontend pages, 8 concurrent AI voice agents, and 8 automation scenarios coordinated through Supabase Edge Functions and VAPI webhooks. This article breaks down the technical architecture: how we designed an 8-table Postgres schema to track agent state, built Edge Functions to handle VAPI webhook callbacks at scale, and packaged the entire system for Vercel deployment. If you're running a multi-tenant voice AI operation or need to coordinate complex agent workflows, this deep dive covers the patterns that worked—from database normalization under concurrent voice traffic to stateless function design for reliable webhook handling.