🛒 DummyCart — Customer Support Resolution Agent

A production-style agentic RAG system that resolves customer support issues end-to-end — answering policy questions, looking up live order data, and escalating to a human when needed.

Features • Architecture • Tech Stack • Getting Started • Demo

---

🚀 Features

• 🧠 Agentic RAG — combines retrieval-augmented generation with tool-calling so the agent grounds every answer in real data
• 📦 Live order lookups — fetches order status, tracking, and item details, gated on customer email verification
• 📄 Knowledge-grounded answers — semantic search over FAQ, refund policy, and shipping policy docs via FAISS
• 🎫 Smart escalation — automatically creates a human support ticket when confidence is low or the customer requests it
• ⚡ Powered by Groq — blazing fast LLM inference using Llama 3.3 70B, completely free
• 🖥️ Streamlit UI + CLI — run as a beautiful chat app or directly in the terminal

---

🏗️ Architecture

User Message
      │
      ▼
┌─────────────────────────────────┐
│     LangGraph ReAct Agent       │
│       (Groq · Llama 3.3 70B)   │
└────────────┬────────────────────┘
             │ picks the right tool
    ┌────────┴─────────┐
    │                  │
    ▼                  ▼
┌─────────────┐  ┌──────────────────┐
│ lookup_order│  │   kb_search /    │
│ orders.json │  │refund_policy_    │
│ email-gated │  │search → FAISS    │
└─────────────┘  └──────────────────┘
    │                  │
    └────────┬─────────┘
             │
             ▼
      ┌─────────────┐
      │create_ticket│  ← escalation
      │tickets.json │
      └─────────────┘
             │
             ▼
      Final Response

---

🧰 The Four Tools

Tool	What it does
lookup_order	Fetches order status, items, totals, tracking — gated on customer email
kb_search	Semantic search across FAQ + all policy docs
refund_policy_search	Targeted search restricted to refund & returns policy
create_ticket	Opens a human support ticket and escalates the conversation

---

🛠️ Tech Stack

Layer	Technology
Agent Framework	LangChain + LangGraph (ReAct agent)
LLM	Groq — Llama 3.3 70B Versatile (free)
Embeddings	HuggingFace — all-MiniLM-L6-v2 (local, free)
Vector Store	FAISS (local)
UI	Streamlit
Language	Python 3.10+

---

📦 Getting Started

Prerequisites

• Python 3.10+
• A free Groq API key — no credit card needed

1. Clone the repo

git clone https://github.com/YOUR_USERNAME/dummycart-support-agent.git
cd dummycart-support-agent

2. Create virtual environment

python -m venv .venv

# Mac/Linux
source .venv/bin/activate

# Windows
.venv\Scripts\Activate.ps1

3. Install dependencies

pip install -e .

4. Configure environment

cp env.example .env

Edit .env and add your Groq API key:

GROQ_API_KEY=your_groq_api_key_here
GROQ_MODEL=llama-3.3-70b-versatile

5. Build the knowledge base (one-time)

python ingest.py

6. Run

Streamlit UI:

streamlit run app.py

Terminal CLI:

python main.py

---

💬 Demo Prompts

📦 Order Lookups

"Check order DC-10234, my email is alex@example.com"
"What's the status of DC-10235? Email: priya@example.com"
"Can you track my order DC-10236? Email: jordan@example.com"

📄 Policy & FAQ

"What's your refund policy?"
"How long does shipping take?"
"Can I cancel my order?"
"What payment methods do you accept?"

🎫 Escalation

"My package arrived broken, I need help"
"I think someone used my card to place an order"
"I want to speak to a human agent"

🔀 Combination

"Where's order DC-10234 (alex@example.com)? Can I still cancel it?"
"What's the return policy for damaged items? My order DC-10236 arrived broken"

---

📂 Project Structure

dummycart-support-agent/
├── agent.py            # LangGraph ReAct agent + system prompt
├── tools.py            # lookup_order, kb_search, refund_policy_search, create_ticket
├── ingest.py           # Builds FAISS index from data/*.md
├── app.py              # Streamlit chat UI
├── main.py             # Terminal CLI
├── data/
│   ├── faq.md
│   ├── refund_policy.md
│   ├── shipping_policy.md
│   ├── orders.json     # Sample order database
│   └── tickets.json    # Tickets created by the agent
├── kb_index/           # FAISS index (auto-created by ingest.py)
├── pyproject.toml
├── env.example
└── .gitignore

---

🔁 Extending It

• Swap data/orders.json for a real database query
• Replace FAISS with a managed vector DB (Qdrant, Pinecone) — only ingest.py changes
• Add tools: update_shipping_address, issue_store_credit, start_return
• Plug create_ticket into Zendesk / Linear / your real ticketing system
• Deploy on Streamlit Cloud for free

---

📄 License

MIT — feel free to use, modify, and build on top of this.

---

Built with ❤️ using LangChain, Groq, and Streamlit