How it Works

A hybrid approach combining deterministic rules with advanced Large Language Models.

⚡

1. Rule Engine

Every ticket first passes through a high-speed Rule Engine. It checks for specific keywords and regex patterns (e.g., "password reset", "error 500").

  • Speed: < 10ms
  • Cost: Free
  • Use Case: Obvious, repetitive requests.
🧠

2. AI Analysis

If no rule matches, the ticket is sent to Google Gemini 2.0 Flash. The AI understands context, sentiment, and nuance to classify complex issues.

  • Speed: ~500ms
  • Intelligence: Understands "My screen is flickering" as a Hardware Issue.
  • Fallback: Automatically switches to OpenAI GPT-4 if Gemini is unavailable.
✅

3. Structured Output

The system returns a standardized JSON response ready for your ticketing system (Jira, Zendesk, etc.).

{ "category": "Hardware Issue", "priority": "High", "confidence": 0.98, "suggested_team": "IT Support" }