I have an AI agent built in n8n that uses webhooks as the entry point, Supabase for session and memory storage, and OpenAI Assistants for responses. The system was working correctly until recent changes to the Supabase data caused runtime failures during webhook-triggered executions. The workflows save and activate correctly, but when a live request hits the webhook, the execution fails due to invalid runtime data (e.g. session/thread handling). I need someone experienced with n8n workflows, webhooks, Supabase, and OpenAI Assistants to trace the root cause, restore the original working behavior, and briefly document what was changed so the system doesn’t break again. All systems are already built. This is a focused debug-and-fix task, not a redesign. Budget is fixed at $25.