Building a Multi-Agent Voice Command Center: Load Bearing Empire's QHV Architecture with VAPI

Description and use

A production voice command center for 8 AI agents across multiple domains, built with VAPI, Supabase Edge Functions, Postgres, and Vercel.

Problem

Scale a production voice AI system to manage multiple concurrent agents across diverse operational domains (valet, real estate, demolition) while maintaining a single backend.

Outcome

The architecture supports 8 concurrent AI agents across multiple domains, providing unified state tracking, automated routing, and a user‑friendly frontend for monitoring and control.

Workflow

• 1. Configure VAPI webhook to trigger Supabase Edge Function on voice input.
• 2. Edge Function processes webhook, updates agent state in Postgres.
• 3. Postgres schema tracks 8 tables: agents, states, scenarios, etc.
• 4. Frontend on Vercel queries Supabase to display agent status and routing options.
• 5. Frontend sends commands back to Edge Function to control agents.
• 6. Edge Function routes commands to appropriate AI agent and updates state.

Evidence

We built a production voice command center handling 8 concurrent AI agents across valet, real estate, and demolition operations—all coordinated through a single Supabase backend.

Source

how we structured VAPI webhook integrations in Supabase Edge Functions, designed an 8-table Postgres schema to track agent states and automation scenarios, and deployed a 5-page frontend on Vercel that manages everything from voice routing to task automation.

Source