🎧 AI-Powered Customer Support Copilot

An intelligent, production-ready support agent that eliminates context-switching for support teams combining RAG, persistent memory, and LangChain tool calling into a one-click response generator.

---

🚀 The Problem

Support agents lose 60%+ of their time switching between CRM, billing systems, and knowledge bases before they can even begin drafting a reply. This project eliminates that entirely.

💡 The Solution

A one-click AI copilot that:

1. Retrieves customer context from CRM and billing systems via tool calling
2. Searches the knowledge base using semantic RAG (ChromaDB)
3. Recalls past interactions using persistent memory (Mem0)
4. Drafts a complete, personalized response ready for agent review in seconds

---

🏗️ Architecture

┌─────────────────────────────────────────────────────┐
│                 Streamlit Dashboard                  │
│         (Ticket Queue | New Ticket | Analytics)      │
└────────────────────┬────────────────────────────────┘
                     │ HTTP
┌────────────────────▼────────────────────────────────┐
│                  FastAPI Backend                     │
│              POST /tickets/{id}/generate             │
└────┬──────────────┬──────────────┬──────────────────┘
     │              │              │
┌────▼────┐  ┌──────▼──────┐  ┌───▼────────────┐
│  Mem0   │  │  ChromaDB   │  │ LangChain Agent│
│ Memory  │  │  RAG Search │  │  + Tool Calling│
└─────────┘  └─────────────┘  └───┬────────────┘
                                   │
                          ┌────────┴────────┐
                          │                 │
                    ┌─────▼─────┐   ┌───────▼───┐
                    │ CRM Tool  │   │Billing Tool│
                    └───────────┘   └───────────┘

🛠️ Tech Stack

Layer	Technology
LLM	OpenAI GPT-4o-mini
Agent Framework	LangChain + OpenAI Tools Agent
Persistent Memory	Mem0
Vector Store	ChromaDB
Backend	FastAPI + SQLAlchemy + SQLite
Frontend	Streamlit
Containerization	Docker + Docker Compose
CI/CD	GitHub Actions → AWS EC2

---

⚡ Quick Start

1. Clone & configure

git clone https://github.com/YOUR_USERNAME/customer-support-copilot.git
cd customer-support-copilot
cp .env.example .env
# Add your OPENAI_API_KEY and MEM0_API_KEY to .env

2. Run with Docker (recommended)

docker compose up --build

• Backend API: http://localhost:8000
• Streamlit UI: http://localhost:8501
• API Docs: http://localhost:8000/docs

3. Run locally (without Docker)

# Backend
cd backend
pip install -r requirements.txt
uvicorn main:app --reload

# Frontend (new terminal)
cd frontend
pip install -r requirements.txt
streamlit run app.py

---

🔑 Required API Keys

Key	Where to Get It
OPENAI_API_KEY	platform.openai.com
MEM0_API_KEY	app.mem0.ai (free tier available)

---

📡 API Endpoints

Method	Endpoint	Description
GET	/health	Health check
GET	/tickets/	List all tickets
POST	/tickets/	Create a new ticket
GET	/tickets/{id}	Get ticket by ID
POST	/tickets/{id}/generate	🤖 Generate AI response
PUT	/tickets/{id}/status	Update ticket status
GET	/customers/	List all customers

Interactive docs available at /docs (Swagger UI).

---

🧠 How the Agent Works

When POST /tickets/{id}/generate is called:

1. RAG Search - Queries ChromaDB for knowledge base articles relevant to the ticket
2. Memory Recall - Fetches past interaction summaries for the customer from Mem0
3. Agent Reasoning - LangChain agent decides which tools to call:
   • get_customer_crm_info(customer_id) - fetches name, plan, account manager
   • get_billing_info(customer_id) - fetches payment status and invoices
4. Response Generation - GPT-4o-mini synthesizes all context into a draft response
5. Memory Update - Interaction is saved to Mem0 for future context

---

🗂️ Project Structure

customer-support-copilot/
├── backend/
│   ├── main.py                  # FastAPI application & routes
│   ├── agent/
│   │   └── support_agent.py     # Core LangChain agent
│   ├── memory/
│   │   └── mem0_client.py       # Mem0 memory client
│   ├── rag/
│   │   └── knowledge_base.py    # ChromaDB vector store
│   ├── tools/
│   │   ├── crm_tool.py          # CRM lookup tool
│   │   └── billing_tool.py      # Billing lookup tool
│   ├── database/
│   │   ├── models.py            # SQLAlchemy ORM models
│   │   └── db.py                # DB init and session
│   └── requirements.txt
├── frontend/
│   └── app.py                   # Streamlit dashboard
├── .github/workflows/
│   └── ci-cd.yml                # GitHub Actions pipeline
├── docker-compose.yml
└── .env.example

---

🚢 CI/CD Pipeline

Push to main
    ↓
Lint (flake8) + Tests
    ↓
Build Docker Images → Docker Hub
    ↓
SSH Deploy → AWS EC2 (docker compose up)

To enable: add DOCKER_USERNAME, DOCKER_PASSWORD, EC2_HOST, EC2_USER, EC2_SSH_KEY as GitHub repository secrets.

---

🙋 Author

Author: Aryan Rajguru