Customer Support Chatbot — Development skill for Claude Code

🎧 Multi-Agent Customer Support Chatbot (AutoGen + Streamlit)

A customer support chatbot built with Microsoft **AutoGen** (`autogen-agentchat` 0.4+) and a **Streamlit** chat UI. A user asks any question or enters a topic. A research agent studies it and sends it to the right department, and that department's specialist agent writes the answer.

How it works

User question
     │
     ▼
🔍 Research & Triage agent  ── researches the question, picks 1 of 5 departments
     │                          (returns research notes, department, confidence, reason)
     ▼
🏢 Department specialist agent ── answers the question and writes helpful content
     │
     ▼
Streamlit chat UI (routing badge + research panel + final answer)

Departments

Department	Handles
💳 Billing	payments, invoices, refunds, pricing, subscriptions
🛠️ Technical Support	bugs, errors, setup, login problems, integrations
📈 Sales	product info, plans, features, demos, upgrades
📦 Shipping & Returns	order tracking, delivery, returns, exchanges
👤 Account & General	account settings, privacy, security, feedback, anything else

Each department reply has four sections: **Answer**, **Step-by-step guidance**, **Helpful content** and **Next steps**.

Features

• Two-stage AutoGen agent pipeline: triage, then a department specialist
• Works with OpenAI (gpt-4o-mini, gpt-4o) or Anthropic (Claude) models, chosen in the sidebar
• Shows which department was picked, the confidence score, and the research notes behind the choice
• Chat history kept for the session, plus a clear-chat button
• If the triage agent's reply can't be read, the question goes to Account & General

Setup

git clone 
cd 
pip install -r requirements.txt
cp .env.example .env      # then add OPENAI_API_KEY and/or ANTHROPIC_API_KEY
streamlit run app.py

You can also paste the API key into the sidebar instead of using `.env`.