Customer Support Agent

An intelligent, AI-powered customer support automation system built with LangChain and LangGraph. Automatically processes, categorizes, prioritizes, and responds to customer complaints with multi-agent workflow orchestration.

Features

• Intelligent Complaint Processing - Automated validation, categorization, and sentiment analysis
• Dynamic Priority Assignment - Smart prioritization based on sentiment, category, and urgency
• Automated Email Responses - Context-aware response generation with policy compliance
• Human Escalation - Automatic escalation for high-priority cases requiring human intervention
• Analytics Dashboard - Real-time metrics, trends, and complaint tracking
• Dual Frontend - Customer complaint submission portal and employee management dashboard
• RESTful API - Complete CRUD operations with FastAPI backend
• Persistent Storage - SQLite database with SQLAlchemy ORM

Architecture

The system uses a multi-agent workflow powered by LangGraph with the following pipeline:

Validation → Categorization → Sentiment Analysis → Priority Assignment 
    → Response Generation → Action Suggestions → Re-evaluation 
    → Escalation Decision → Human Review (if needed)

Agent Pipeline

1. Validation Agent - Validates complaint legitimacy and completeness
2. Categorization Agent - Classifies complaints into predefined categories
3. Sentiment Agent - Analyzes emotional tone (positive, negative, neutral)
4. Priority Agent - Assigns priority levels (low, medium, high, critical)
5. Response Agent - Generates appropriate email responses
6. Suggestion Agent - Recommends actionable steps for resolution
7. Re-evaluation Agent - Reviews priority based on context
8. Escalation Agent - Determines if human intervention is required

Tech Stack

Backend:

• Python 3.x
• FastAPI - REST API framework
• LangChain - LLM orchestration
• LangGraph - Multi-agent workflow
• Groq - LLM provider
• SQLAlchemy - ORM
• SQLite - Database
• Pydantic - Data validation

Frontend:

• React - Customer portal
• React - Employee dashboard
• Axios - HTTP client

Project Structure

Customer-Support-Agent/
├── Agent/
│   ├── agents/          # Individual agent implementations
│   ├── memory/          # Conversation memory management
│   ├── policies/        # Business rules and policies
│   ├── schemas/         # Pydantic models
│   ├── utils/           # Helper utilities
│   ├── graph.py         # LangGraph workflow definition
│   ├── state.py         # State management
│   └── main.py          # Agent entry point
├── Backend/
│   ├── api.py           # FastAPI routes
│   ├── crud.py          # Database operations
│   ├── models.py        # SQLAlchemy models
│   ├── schemas.py       # Pydantic schemas
│   ├── worker.py        # Background worker
│   └── db.py            # Database configuration
├── Frontend/
│   ├── customer/        # Customer complaint portal
│   └── employee-dashboard/  # Staff management interface
├── Email/               # Email notification system
└── FinalisingAgent/     # Final processing logic

Getting Started

Prerequisites

• Python 3.8+
• Node.js 16+ (for frontend)
• Groq API key

Installation

1. Clone the repository

git clone https://github.com/MasirJafri1/Customer-Support-Agent.git
cd Customer-Support-Agent

2. Install Python dependencies

pip install -r requirements.txt

3. Configure environment variables

Create a .env file:

GROQ_API_KEY=your_groq_api_key
DATABASE_URL=sqlite:///./complaints.db

4. Initialize the database

python seed_data.py

5. Start the FastAPI backend

python main_api.py
# or
uvicorn Backend.api:app --reload

6. Start the background worker

python main_worker.py

7. Start the frontend (in separate terminals)

# Customer portal
cd Frontend/customer
npm install
npm run dev

# Employee dashboard
cd Frontend/employee-dashboard
npm install
npm run dev

API Endpoints

Complaints

• POST /complaints - Submit a new complaint
• GET /complaints - List all complaints (with optional filters)
• GET /complaints/{id} - Get complaint by ID
• PATCH /complaints/{id} - Update complaint status/priority
• DELETE /complaints/{id} - Delete a complaint
• GET /complaints/by-email/{email} - Get complaints for a customer

Analytics

• GET /analytics/summary - Overall metrics summary
• GET /analytics/status - Breakdown by status
• GET /analytics/priority - Breakdown by priority
• GET /analytics/trends - Daily complaint trends

Usage

Customer Flow

1. Visit the customer portal
2. Enter email, order ID, and complaint details
3. System automatically processes and responds
4. Receive automated email response

Employee Flow

1. Access employee dashboard
2. View all complaints with filters
3. Monitor real-time analytics
4. Handle escalated cases requiring human review

Key Features Explained

Intelligent Categorization

Complaints are automatically classified into categories like:

• Product Quality
• Shipping Issues
• Customer Service
• Billing Problems
• Technical Issues

Priority System

• Low - Minor issues, no immediate action needed
• Medium - Moderate concerns, standard resolution timeline
• High - Significant issues requiring prompt attention
• Critical - Urgent matters with potential escalation

Automated Responses

AI-generated responses are:

• Contextually relevant
• Policy-compliant
• Empathetic and professional
• Actionable with clear next steps

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License.

Author

Masir Jafri

• Website: masirjafri.in
• GitHub: @MasirJafri1
• LinkedIn: masirjafri

Acknowledgments

• Built with LangChain and LangGraph
• Powered by Groq
• API framework: FastAPI

---