🚀 Just shipped: BFS Suite (Big For Small) — a multi-agent AI support system built entirely with n8n + RAG

29d (edited) • Wins

Here's what I built:

3 specialized AI agents under one roof:

→ Legal agent — contracts & compliance

→ HR agent — policies & onboarding

→ Support agent — FAQs & troubleshooting

Every query goes through an AI Router (GPT-4o-mini) that classifies the topic and routes to the right agent in real time.

⚙️ Under the hood:

• RAG with pgvector similarity search (Supabase)

• Dual access: public web widget vs internal dashboard

• Knowledge gap detection — unanswered questions logged automatically

• Escalation via Telegram — 3 triggers: low similarity score, explicit user request, or ambiguous routing (Legal/HR overlap)

• Human-in-the-loop — when the router's confidence drops below 0.7, the query goes to a human instead of guessing

• Full analytics dashboard with live charts

• Document upload straight from the dashboard

🛠 Stack: n8n · GPT-4o-mini · Supabase (pgvector) · Netlify · Telegram Bot API

📊 5 workflows: Ingestion → Query → Logging → Upload → Escalation

Everything is production-ready — real RAG, real logging, real escalation. Not a toy project.

Happy to answer any questions about the build — drop them below 👇

4 comments

🚀 Just shipped: BFS Suite (Big For Small) — a multi-agent AI support system built entirely with n8n + RAG

skool.com/data-and-ai

For operators and builders who want to actually profit from AI — not just learn about it.

Leaderboard (30-day)