Building a Multi-Agent Voice Automation Command Center: Load Bearing Empire's QHV Stack (Supabase + VAPI + Vercel)

OPENING: We built Quiet Hours Valet's command center from scratch in 2026—a full-stack voice automation platform handling 8 concurrent AI agents, 8 automation workflows, and real-time dispatch across a trash valet operation. Here's the technical breakdown: Supabase Edge Functions handling VAPI webhook ingestion, a normalized 8-table Postgres schema for agent state management, 5 responsive frontend pages deployed on Vercel, and self-hosted Asterisk PBX integration for phone routing. If you're building polyglot AI systems that need to scale beyond chatbots into actual business operations, this stack handles order fulfillment, customer routing, callback queuing, and agent orchestration without the SaaS tax. We'll walk through the architecture decisions, the Postgres schema that made webhook processing sane, and why Edge Functions beat traditional serverless for latency-critical voice automation.