🎫 AI Ticket Assistant: Next-Gen AI-Powered Ticketing System

AI Ticket Assistant is a production-ready, AI-native customer support platform. Automating ticket triage, skill-based routing, and response drafting using Groq LLM (LLaMA 3.3 70B).

Decoupled AI and email workflows using Inngest event-driven architecture, keeping Express response times at 20–50ms. Optimized MongoDB queries via field projection and selective .populate() to prevent N+1 issues and reduce payload size. Secured with JWT authentication, role-based access control, Helmet, rate limiting, and Zod validation.

---

🚀 Key Features

• 🧠 AI-Powered Triage: Every incoming ticket is automatically analyzed by Groq LLaMA 3.3 for priority (low/medium/high), helpful notes for moderators, and required technical skills.
• ⚡ Smart AI-Drafted Replies: Staff can generate high-quality, context-aware reply drafts with a single click via Groq Mixtral, drastically reducing resolution time.
• 🛠️ Skill-Based Routing: Tickets are automatically matched and assigned to moderators based on their specific technical expertise (React, Node.js, etc.).
• 🏗️ Reliable Event-Driven Architecture: Powered by Inngest for guaranteed background job execution, automatic retries, and real-time observability — each pipeline step is independently retried on failure.
• 🔒 Secure Role-Based Access Control: Granular permissions for user, moderator, and admin roles.
• 📋 Service Layer Architecture: Clean separation of concerns — controllers handle HTTP, services handle business logic, models handle persistence.
• 🗑️ Soft Deletes: Tickets and users are never permanently destroyed — only flagged as deleted.
• 📄 Interactive API Docs: A dark-themed documentation page served at / showcasing all endpoints with real request/response examples.

---

🛠️ Tech Stack

Layer	Technology
Runtime	Node.js v18+ (ESM)
Framework	Express.js
Database	MongoDB + Mongoose (Atlas)
Auth	JSON Web Tokens (JWT)
Async Workflows	Inngest
AI / LLM	Groq API (LLaMA 3.3 70B, Mixtral 8x7B)
Validation	Zod
Security	Helmet, express-rate-limit, bcrypt
Logging	Winston + Morgan
Email	Nodemailer + Mailtrap

---

📁 Project Structure

InngestKit-TicketProject/
├── AI-ticket-Assistant/        # Node.js + Express Backend
│   ├── config/config.js        # Centralized environment config
│   ├── controllers/            # HTTP request/response handlers
│   ├── services/               # Business logic layer
│   ├── models/                 # Mongoose schemas
│   ├── routes/                 # Express route definitions
│   ├── middlewares/            # Auth + Zod validation
│   ├── inngest/                # Background workflow functions
│   │   └── inngest-Functions/
│   │       ├── on-signup.js         # Welcome email on registration
│   │       └── on-ticketCreated.js  # Full 6-step AI triage pipeline
│   ├── validations/            # Zod schemas
│   ├── utils/                  # Logger, mailer, Groq AI helpers
│   └── public/index.html       # Interactive API documentation UI

---

⚙️ Local Setup

Prerequisites

• Node.js v18+
• MongoDB Atlas account (or local MongoDB)
• Inngest CLI for local workflow testing
• Groq API key (free at groq.com)
• Mailtrap account for email testing

1. Clone the repository

git clone https://github.com/YOUR-USERNAME/AI-ticket-Project.git
cd AI-ticket-Project/AI-ticket-Assistant

2. Install dependencies

npm install

3. Set up environment variables

Create a .env file based on .env.example:

PORT=4000
NODE_ENV=development
MONGO_URI=mongodb+srv://...
JWT_SECRET=your_super_secret_key
GROQ_API_KEY=gsk_...
INNGEST_EVENT_KEY=local
ALLOWED_ORIGIN=http://localhost:3000
MAILTRAP_SMTP_USER=your_mailtrap_user
MAILTRAP_SMTP_PASS=your_mailtrap_pass

4. Run locally

# Terminal 1 — Express API
npm run start

# Terminal 2 — Inngest Dev Server (for background workflow testing)
npx inngest-cli@latest dev

Visit http://localhost:4000/ for the interactive API docs.

---

📡 API Endpoints

Auth — /api/auth

Method	Endpoint	Access	Description
POST	/signup	Public	Register a new user
POST	/login	Public	Login and receive JWT
GET	/details	🔒 Auth	Get current user profile
PUT	/update	⭐ Admin	Update any user's role and skills
PUT	/profile	🔒 Auth	Update current user's skills/avatar

Tickets — /api/tickets

Method	Endpoint	Access	Description
POST	/	🔒 Auth	Create a ticket — triggers AI triage
GET	/?page=1&limit=10	🔒 Auth	Get paginated tickets
GET	/:id	🔒 Auth	Get a single ticket with responses
POST	/:id/responses	🔒 Auth	Add a reply to a ticket thread
POST	/:id/draft	⭐ Mod/Admin	Generate an AI reply draft

System

Method	Endpoint	Access	Description
GET	/	Public	Interactive API documentation
GET	/health	Public	Server & database health check

---

🤖 Inngest AI Pipeline

When a ticket is created, a durable 6-step Inngest workflow runs asynchronously:

1. fetch-ticket            → Load ticket from MongoDB
2. update-ticket-status    → Set status → TODO
3. call-ai-triage          → Groq LLaMA 3.3: analyze priority, notes & skills
4. ai-processing           → Save AI results, status → IN_PROGRESS
5. assign-moderator        → Regex match moderator by relatedSkills
6. send-email-notification → Email the assigned moderator via Mailtrap

Each step is independently retried on failure (max 2 retries) — the entire pipeline never restarts from scratch.

---

🔐 Role Permissions

Action	user	moderator	admin
Create ticket	✅	✅	✅
View own tickets	✅	✅	✅
View all tickets	❌	✅	✅
Add response	✅	✅	✅
Generate AI draft	❌	✅	✅
Update user roles	❌	❌	✅

---

Built with ❤️ — showcasing AI agents handling complex, multi-step business workflows from categorization to automated assignment and human-in-the-loop AI assistance.