Building a Multi-Agent Voice Command Center: Load Bearing Empire's QHV Architecture with VAPI, Supabase Edge Functions, and Verc

OPENING PARAGRAPH:

We built Quiet Hours Valet's command center in 72 hours—a production voice operations platform handling 8 concurrent AI agents, 8 automation workflows, and real-time dispatch across a fleet of valet services. This article breaks down the technical stack we chose: VAPI for voice inference, Supabase Edge Functions for webhook processing at scale, a normalized 8-table Postgres schema for agent state management, Asterisk PBX for legacy integration, and Vercel for deployment. If you're scaling voice AI beyond proof-of-concept—multi-tenant SaaS, field operations, or compliance-heavy workflows—this is a battle-tested architecture that handles concurrent agent invocations, webhook reliability, and stateful automation without serverless timeouts or cold starts.