🎫 AI Ticket Classifier

A lightweight, production-ready Python library for automatically classifying support tickets using pattern matching and optional LLM integration.

✨ Features

• 🚀 Zero dependencies for basic classification (pattern-based)
• 🤖 Optional LLM support (OpenAI, Azure OpenAI, local LLMs)
• 📦 Pre-configured categories for common IT support scenarios
• 🎯 High accuracy with confidence scoring
• 🔧 Fully customizable - add your own categories and patterns
• ⚡ Fast - classify thousands of tickets per second (pattern mode)
• 🔌 Platform agnostic - works with any ticketing system

📦 Installation

Basic Installation (Pattern Matching Only)

pip install ai-ticket-classifier

With LLM Support

pip install ai-ticket-classifier[llm]
# or
pip install openai requests

🚀 Quick Start

Basic Pattern-Based Classification

from ticket_classifier import TicketClassifier

# Create classifier
classifier = TicketClassifier()

# Classify a ticket
ticket = "I forgot my password and can't log in"
result = classifier.classify(ticket)

print(f"Category: {result.category.name}")
print(f"Confidence: {result.confidence:.2%}")
print(f"Priority: {result.category.priority}")
print(f"Auto-resolvable: {result.category.auto_resolvable}")

Output:

Category: password_reset
Confidence: 95%
Priority: high
Auto-resolvable: True

LLM-Based Classification (More Accurate)

from ticket_classifier import LLMClassifier

# Using OpenAI
classifier = LLMClassifier(
    api_key="your-openai-key",
    provider="openai",
    model="gpt-3.5-turbo"
)

# Or use local LLM (LM Studio, Ollama)
classifier = LLMClassifier(
    provider="local",
    api_base="http://127.0.0.1:1234/v1"
)

result = classifier.classify("My computer is really slow")
print(result.category.name)  # application_error

Batch Classification

tickets = [
    "Need password reset",
    "Printer not working",
    "C drive full"
]

results = classifier.classify_batch(tickets)

for ticket, result in zip(tickets, results):
    print(f"{ticket} → {result.category.name}")

📋 Pre-configured Categories

The library includes 11 common IT support categories:

Category	Description	Auto-Resolvable
password_reset	Password resets and account unlocks	✅ Yes
disk_space	Low disk space issues	✅ Yes
printer_issue	Printer problems	✅ Yes
email_issue	Email sending/receiving problems	❌ No
software_install	Software installation requests	❌ No
network_issue	Network/VPN connectivity	❌ No
license_request	Software license requests	✅ Yes
hardware_issue	Hardware malfunctions	❌ No
access_request	File/folder access requests	✅ Yes
application_error	Application crashes/errors	❌ No
other	Uncategorized issues	❌ No

🎨 Custom Categories

Add your own categories:

from ticket_classifier import TicketClassifier, TicketCategory

classifier = TicketClassifier()

# Add custom category
custom_category = TicketCategory(
    name="voip_issue",
    description="VoIP phone system problems",
    keywords=["phone", "voip", "call", "extension"],
    patterns=[
        r"phone.*not.*work",
        r"no.*dial.*tone",
        r"can'?t.*make.*call"
    ],
    priority="high",
    auto_resolvable=False
)

classifier.add_category(custom_category)

🔌 Integration Examples

With Zendesk

from ticket_classifier import TicketClassifier
from zenpy import Zenpy

classifier = TicketClassifier()
zendesk = Zenpy(**creds)

for ticket in zendesk.tickets():
    result = classifier.classify(ticket.subject + " " + ticket.description)

    # Add tags
    ticket.tags.extend([result.category.name, result.category.priority])

    # Set priority
    ticket.priority = result.category.priority

    zendesk.tickets.update(ticket)

With Atera

import requests
from ticket_classifier import TicketClassifier

classifier = TicketClassifier()

# Get tickets from Atera
response = requests.get(
    "https://app.atera.com/api/v3/tickets",
    headers={"X-API-KEY": "your-key"}
)

for ticket in response.json():
    result = classifier.classify(ticket['Title'])

    # Update ticket with classification
    requests.put(
        f"https://app.atera.com/api/v3/tickets/{ticket['TicketID']}",
        headers={"X-API-KEY": "your-key"},
        json={"Comment": f"Auto-classified as: {result.category.name}"}
    )

With Zoho Desk

from ticket_classifier import TicketClassifier

classifier = TicketClassifier()

# Classify incoming webhook
def classify_webhook(ticket_data):
    result = classifier.classify(ticket_data['subject'])

    return {
        "category": result.category.name,
        "priority": result.category.priority,
        "auto_resolve": result.category.auto_resolvable
    }

🎯 Use Cases

• Automated Triage: Route tickets to the right team automatically
• Priority Assignment: Set ticket priority based on classification
• Auto-Resolution: Identify tickets that can be auto-resolved
• Analytics: Track ticket categories and trends
• SLA Management: Different SLAs for different ticket types
• Chatbot Training: Use classifications to train support chatbots

📊 Performance

Pattern-based classifier:

• ~10,000 tickets/second
• Zero API costs
• No external dependencies

LLM-based classifier:

• More accurate for complex/nuanced tickets
• ~10-50 tickets/second (depending on LLM provider)
• API costs apply

🤝 Contributing

Contributions welcome! Areas where you can help:

• Add more default categories
• Improve pattern matching accuracy
• Add support for more languages
• Create integrations for popular ticketing systems

📄 License

MIT License - see LICENSE file

🏢 Commercial Support

This is an open-source project by Turtles AI Lab.

Need help integrating this into your business?

• Custom category development
• Integration consulting
• Production deployment support

📧 Contact: jgreenia@jandraisolutions.com

🌟 Related Projects

• TicketZero AI - Atera Edition - Full automated ticket resolution for Atera
• TicketZero AI - Zoho Edition - Full automated ticket resolution for Zoho Desk

---

Built with ❤️ by Turtles AI Lab

⭐ Star this repo if you find it useful!