🎫 Customer Support Ticket Classifier

An AI-powered classification and routing system that utilizes Semantic Similarity Search to automatically categorize support tickets and direct them to the appropriate departments.

🚀 Overview

Traditional ticket systems rely on rigid, keyword-based rules that often fail to understand context. This system uses Transformer-based embeddings to understand the meaning behind a customer's inquiry. By comparing new tickets against historical data in the Endee Vector Database, it provides high-accuracy routing, priority assignment, and SLA tracking.

Key Features

• 🤖 Automatic Classification: Uses sentence-transformers to generate 384-dimensional semantic embeddings.
• 🎯 Smart Routing: Automatically maps tickets to departments (Technical, Billing, Engineering, etc.).
• 📈 Confidence Scoring: Provides a reliability score for every classification to ensure high-quality routing.
• 🔌 REST API: Fully documented endpoints for ticket submission and management.
• 🖥️ Admin Dashboard: Integrated Django admin interface for monitoring and manual overrides.
• 📱 Modern UI: Responsive web interface for seamless ticket submission.

---

🛠️ Tech Stack

Component	Technology
Backend	Python 3.8+, Django, Django REST Framework
Vector Database	Endee (Running via Docker)
ML Models	Sentence-Transformers (all-MiniLM-L6-v2)
Primary Database	SQLite (Development)
Frontend	HTML5, CSS3, JavaScript

---

🏗️ Project Structure

ticket-classifier/
├── core/                      # ML & Vector Logic
│   ├── classifier.py          # Ticket classification engine
│   ├── embedding_service.py   # Vector generation
│   └── endee_client.py        # Endee API integration
│
├── tickets/                   # Django Application
│   ├── models.py              # Ticket & Category schemas
│   ├── views.py               # API & Web views
│   └── serializers.py         # DRF Serializers
│
├── scripts/                   # Data Pipeline
│   ├── preprocess_data.py     # CSV Cleaning & Splitting
│   └── load_to_endee.py       # Vector Ingestion
│
└── config.py                  # Environment Configuration

---

🚦 Quick Start

1. Prerequisites

• Python 3.8+
• Docker (Required to run the Endee Vector Database)

2. Start Endee Vector Engine

# Start the container
docker start endee-server

# Verify status
docker ps --filter name=endee-server

3. Setup Environment

# Clone the repository
git clone [https://github.com/your-username/ticket-classifier.git](https://github.com/your-username/ticket-classifier.git)
cd ticket-classifier

# Create and activate virtual environment
python -m venv venv
.\venv\Scripts\Activate.ps1  # Windows
source venv/bin/activate     # Mac/Linux

# Install dependencies
pip install -r requirements.txt

4. Initialize & Run

# Apply migrations
python manage.py migrate

# Create admin user
python manage.py createsuperuser

# Start server
python manage.py runserver

---

🔌 API Reference

Submit a Ticket

POST /api/tickets/submit/

Request:

{
  "customer_name": "John Doe",
  "customer_email": "john@example.com",
  "subject": "Cannot access my account",
  "description": "I forgot my password and the reset email isn't arriving"
}

Response:

{
  "ticket": {
    "ticket_id": "uuid-12345",
    "predicted_category": "Account Access",
    "confidence_score": 0.85,
    "department": "Account Services",
    "priority": "high",
    "sla_hours": 2,
    "status": "new"
  }
}

---

🧠 How It Works

1. Vectorization: The system combines subject and description and converts the text into a numerical vector using all-MiniLM-L6-v2.
2. Semantic Search: The vector is queried against the Endee Vector Database to find the most similar historical tickets.
3. Weighted Voting: The system predicts the category based on the labels of the nearest neighbors, weighted by their similarity score.
4. Routing Engine: Based on the predicted category, the system assigns a Department, Priority level, and SLA.

---

📊 Routing Matrix

Category	Department	Priority	SLA (Hours)
Technical	Technical Support	High	4
Account	Account Services	High	2
Billing	Billing Team	Medium	24
Bug	Engineering	High	8
Product	Sales Team	Medium	8

---

📝 License

This project is licensed under the Apache License 2.0.

---

Author:Samarth Umrao