support-agent

Support Agent

Phase 4 capstone — GenAI Full-Stack Developer Track

Customer support multi-agent system — one app combining a support chatbot with specialist agents orchestrated by a supervisor.

Stack: Next.js 15 + FastAPI + OpenAI (or Ollama)

---

What you build

A realistic AcmeShop support desk where customer messages flow through multiple agents:

Customer message
    → Triage Agent (classify intent)
    → Supervisor (route)
    → Knowledge | Orders | Escalation agent
    → Supervisor (polish reply)
    → Customer-facing answer + agent trace

Agent	Role
Triage	Classify intent (returns, billing, order status, …)
Knowledge	Answer from support KB (returns, shipping, billing policies)
Orders	Look up mock order data (ORD-1001, …)
Escalation	Create human handoff tickets
Supervisor	Route + polish final response

---

Quick start

cp backend/.env.example backend/.env    # add OPENAI_API_KEY
make install
make dev

• Support UI: http://localhost:3001
• API docs: http://127.0.0.1:8001/docs

Uses ports 3001 / 8001 so it can run alongside rag-assistant.

---

Try these prompts

Prompt	Expected routing
"What's your refund policy?"	triage → knowledge
"Where is order ORD-1001?"	triage → orders
"I was charged twice"	triage → knowledge (billing)
"I want to speak to a human"	triage → escalation

Expand the Agent trace panel on each reply to see the multi-agent workflow.

---

Learning docs

Start here: LEARNING-GUIDE.md

Doc	Topic
01-agents-and-multi-agent	Agents vs RAG, multi-agent patterns
02-architecture	System design
03-setup-and-run	Install & troubleshooting

---

Project layout

support-agent/
├── LEARNING-GUIDE.md
├── docs/
├── sample-data/
│   ├── knowledge-base.md    # Support policies
│   └── orders.json          # Mock orders
├── backend/
│   ├── app/agents/          # Triage, Knowledge, Orders, Escalation
│   ├── app/services/        # Orchestrator, LLM, stores
│   └── tests/
├── frontend/                # Chat UI + agent trace
└── Makefile