Description
Summary There is an AI platform running in production on top of GoHighLevel for the agency and its sub-accounts. It powers automated conversations (SMS, email, web chat, FB/IG, voice), lead qualification, appointment booking, and knowledge-base-grounded (RAG) responses across a multi-tenant footprint. It is structurally sound and generally runs fine. But now and then there are intermittent issues - a duplicate reply, a booking an hour off, an answer that ignores the knowledge base - and a few have escalated into serious problems (double-booked calendars, a sub-account silently disconnecting, an answer that looked like it came from the wrong client's data). I want an expert who can build the complete, perfect structure for the GHL AI platform. You will: - Review the architecture doc (later, the codebase). - Diagnose the subtle root causes behind the intermittent-but-serious issues. - Deliver a completed, production-grade architecture, plus a prioritized remediation plan. Required: - GHL API v2, OAuth (marketplace app), and webhooks - real-world quirks included. - Distributed systems: idempotency, delivery semantics, message ordering, race conditions. - Multi-tenant isolation, RAG/vector search, LLM context & tool-use, time zones/DST, rate limiting. Preferred: GHL/conversation-AI at scale; HIPAA/PII systems; Claude tool-use.