Omnichannel
Shared logic for WhatsApp and email.
MINE MProject 04 · Omnichannel AI agent
Asteria receives WhatsApp and email messages, answers from approved knowledge, qualifies needs, proposes appointments and hands sensitive decisions to a person.
Understand in 30 seconds
An inbound message is normalized and stored. Rules and a model classify it. The agent can retrieve approved knowledge, continue a commercial conversation or prepare an appointment. Complaints, refunds and sensitive situations create a human task.
sanitized and documented n8n exports
WhatsApp Business and Gmail
reproducible Supabase foundations
Figures verified in the public repository. No client outcome or production load result is claimed.
01 — Journey
Every stage leaves an audit trail and follows explicit limits.
Both formats become one common structure.
The system chooses automation or human handoff.
RAG answers with sources; Calendar acts after confirmation.
A task preserves the escalation reason and context.
02 — Capabilities
Shared logic for WhatsApp and email.
Facts come from approved documents.
The agent understands needs and presents services.
Proposal, confirmation and Calendar creation.
Sensitive requests leave the autonomous journey.
Supabase stores decisions, statuses and events.
03 — Technologies
n8n · OpenAI · deterministic rules · RAG
WhatsApp Business Cloud API · Gmail · Google Calendar
Supabase/PostgreSQL · Next.js
Transparency
Asteria is a personal prototype using a fictional company and fictional data. It is not a customer-service system deployed in production. Security, monitoring, failure recovery and load testing would need strengthening before real-world use.