Building a Multi-Agent Valet Trash Command Center: Supabase Edge Functions, VAPI, and Postgres at Scale

Description and use

A unified command center was built to manage 8 AI voice agents, 5 frontend pages, and 8 automation workflows using Supabase Edge Functions, VAPI webhooks, and a normalized Postgres schema, enabling scalable real‑time coordination and seamless deployment via Vercel.

Problem

Managing 8 concurrent AI voice agents, 5 frontend pages, and 8 discrete automation workflows for a waste services operation required a scalable, real‑time coordination system that could ingest high‑frequency webhook events without disrupting live voice calls.

Outcome

The command center supports 8 concurrent AI voice agents, 5 frontend pages, and 8 automation workflows, handling high‑frequency webhook ingestion with efficient database indexing and real‑time coordination, while enabling seamless updates via Vercel.

Workflow

• 1. VAPI webhook receives voice event
• 2. Supabase Edge Function processes webhook and updates Postgres
• 3. Edge Function triggers appropriate AI voice agent workflow
• 4. Agent processes request and returns response
• 5. Response sent back through VAPI to caller
• 6. Frontend pages display status and logs

Evidence

We built a production command center for Quiet Hours Valet handling 8 concurrent AI voice agents, 5 frontend pages, and 8 discrete automation workflows—all orchestrated through Supabase Edge Functions triggering VAPI webhooks and a normalized 8-table Postgres schema.

Source

This technical deep-dive covers the architecture decisions behind a vertically integrated waste services operation running on self-hosted infrastructure, how we structured real-time voice agent coordination through serverless functions, database indexing strategies for high-frequency webhook ingestion, and the deployment pipeline we use to ship updates to Vercel without disrupting live voice calls.

Source