AI Multi-Agent Token Triage System

Intelligent Support Ticket Management with Autonomous AI Agents

An AI-powered support ticket system where autonomous agents handle the entire triage workflow — from analyzing issues and extracting priorities to matching tickets with the right moderators and sending notifications. No manual routing needed.

---

Demo

TODO: Add screenshots and demo video. See the Media section for recommended assets.

---

Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                          FRONTEND (React + Vite)                    │
│                                                                     │
│   ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────────────┐  │
│   │  Signup   │  │  Login   │  │  Tokens  │  │   Admin Panel    │  │
│   │  Page     │  │  Page    │  │  Dashboard│  │   (User Mgmt)    │  │
│   └──────────┘  └──────────┘  └──────────┘  └──────────────────┘  │
│          │             │             │               │              │
│          └─────────────┴─────────────┴───────────────┘              │
│                              │ REST API                             │
└──────────────────────────────┼──────────────────────────────────────┘
                               │
┌──────────────────────────────┼──────────────────────────────────────┐
│                    BACKEND (Express + MongoDB)                      │
│                              │                                      │
│   ┌──────────────────────────▼───────────────────────────────┐     │
│   │                    Express API Server                      │     │
│   │  /api/auth/*          /api/tokens/*        /api/inngest   │     │
│   └──────────────────────────┬───────────────────────────────┘     │
│                              │                                      │
│          ┌───────────────────┼───────────────────┐                  │
│          │                   │                   │                  │
│          ▼                   ▼                   ▼                  │
│   ┌─────────────┐    ┌─────────────┐    ┌─────────────┐            │
│   │   MongoDB   │    │   Inngest   │    │  Mailtrap   │            │
│   │  Database   │    │  Workflows  │    │    SMTP     │            │
│   └─────────────┘    └──────┬──────┘    └─────────────┘            │
│                              │                                      │
│         ┌────────────────────┼────────────────────┐                 │
│         │                    │                    │                 │
│         ▼                    ▼                    ▼                 │
│   ┌───────────┐     ┌───────────────┐     ┌──────────────┐         │
│   │  Agent 1  │     │    Agent 2    │     │   Google     │         │
│   │  Welcome  │     │  AI Triage    │     │   Gemini     │         │
│   │  Email    │     │  Pipeline     │     │   2.5 Flash  │         │
│   └───────────┘     └───────────────┘     └──────────────┘         │
│                                                                     │
└─────────────────────────────────────────────────────────────────────┘

How It Works

This system uses two autonomous AI agents built on Inngest's durable workflow engine:

Agent 1: Welcome Email Agent

Triggered when a new user signs up. Looks up the user in the database and sends a welcome email via Mailtrap SMTP.

Agent 2: AI Triage Agent (Core)

Triggered when a support token is created. Runs a 6-step pipeline:

Token Created
     │
     ▼
┌─────────────────┐
│ 1. Fetch Token  │  Retrieve from MongoDB
└────────┬────────┘
         ▼
┌─────────────────┐
│ 2. Set Status   │  Mark as "TODO"
└────────┬────────┘
         ▼
┌─────────────────────────────────────────┐
│ 3. AI Analysis (Google Gemini 2.5 Flash) │
│    • Summary extraction                  │
│    • Priority classification             │
│    • Helpful notes generation            │
│    • Related skills extraction           │
└────────┬────────────────────────────────┘
         ▼
┌─────────────────┐
│ 4. Update Token │  Save AI results to DB
└────────┬────────┘
         ▼
┌─────────────────────────────────────────┐
│ 5. Auto-Assignment                       │
│    • Match skills → moderator            │
│    • Fallback → admin                    │
└────────┬────────────────────────────────┘
         ▼
┌─────────────────┐
│ 6. Notify       │  Email assigned moderator
└─────────────────┘

Key Features

• Autonomous AI Triage — Google Gemini analyzes each ticket, extracts priority, summary, and required skills automatically
• Smart Assignment — Tickets are routed to moderators based on skill matching, with admin fallback
• Durable Workflows — Inngest ensures each pipeline step is retriable and fault-tolerant
• Role-Based Access — Three roles (user, moderator, admin) with distinct permissions
• Skill-Based Routing — Moderators have skill tags; AI matches ticket requirements to the right person
• Email Notifications — Assigned moderators receive automatic email alerts
• Admin Panel — Full user management: roles, skills, and search

Tech Stack

Layer	Technology	Purpose
AI Engine	Google Gemini 2.5 Flash	Ticket analysis, priority extraction, skill matching
Agent Framework	Inngest + @inngest/agent-kit	Durable multi-step workflows with retries
Backend	Express v5 + Node.js	REST API server
Database	MongoDB + Mongoose	User and token persistence
Auth	JWT + bcrypt	Token-based authentication with password hashing
Frontend	React 19 + Vite	Single-page application
Styling	Tailwind CSS v4 + DaisyUI v5	UI components and responsive design
Email	Nodemailer + Mailtrap	SMTP email delivery (dev sandbox)

Project Structure

AI-Multi-Agents-v1/
├── ai-token-assistant/              # Backend API
│   ├── index.js                     # Express server entry point
│   ├── env.js                       # Environment config loader
│   ├── makeAdmin.js                 # CLI: promote user to admin
│   ├── controllers/
│   │   ├── token.js                 # Token CRUD + creation event
│   │   └── user.js                  # Auth + user management
│   ├── inngest/
│   │   ├── client.js                # Inngest client setup
│   │   └── functions/
│   │       ├── onsignup.js          # Agent 1: welcome email
│   │       └── ont-token-create.js  # Agent 2: AI triage pipeline
│   ├── middlewares/
│   │   └── auth.js                  # JWT verification + role guards
│   ├── models/
│   │   ├── token.js                 # Token schema (title, priority, skills...)
│   │   └── user.js                  # User schema (email, role, skills)
│   ├── routes/
│   │   ├── token.js                 # /api/tokens/*
│   │   └── user.js                  # /api/auth/*
│   ├── scripts/
│   │   └── cleanDb.js               # Database cleanup utility
│   └── utils/
│       ├── Ai.js                    # Gemini AI agent wrapper
│       └── mailer.js                # Nodemailer SMTP config
│
├── ai-token-frontend/               # React SPA
│   ├── src/
│   │   ├── main.jsx                 # App entry point
│   │   ├── components/
│   │   │   ├── checkauth.jsx        # Route guard (auth protection)
│   │   │   └── navbar.jsx           # Dynamic navigation bar
│   │   ├── lib/
│   │   │   └── api.js               # API client (axios)
│   │   └── pages/
│   │       ├── login.jsx            # Login page
│   │       ├── Signup.jsx           # Registration page
│   │       ├── tokens.jsx           # Token list + creation
│   │       ├── token.jsx            # Token detail view
│   │       └── admin.jsx            # Admin user management
│   └── vite.config.js
│
└── README.md

Getting Started

Prerequisites

• Node.js 18+
• MongoDB instance (local or Atlas)
• Google Gemini API key
• Mailtrap account (for email testing)

Installation

# Clone the repository
git clone https://github.com/your-username/AI-Multi-Agents-v1.git
cd AI-Multi-Agents-v1

Backend setup:

cd ai-token-assistant
npm install

# Create environment file
cp .env.sample .env
# Edit .env with your credentials

Frontend setup:

cd ../ai-token-frontend
npm install

# Create environment file
cp .env.example .env
# Set VITE_SERVER_URL=http://localhost:3000/api

Environment Variables

Backend (ai-token-assistant/.env):

Variable	Description
MONGO_URI	MongoDB connection string
JWT_SECRET	Secret key for JWT signing
GEMINI_API_KEY	Google Gemini API key
MAILTRAP_SMTP_HOST	Mailtrap SMTP host
MAILTRAP_SMTP_PORT	Mailtrap SMTP port
MAILTRAP_SMTP_USER	Mailtrap SMTP username
MAILTRAP_SMTP_PASS	Mailtrap SMTP password
APP_URL	Frontend URL (http://localhost:5173)

Frontend (ai-token-frontend/.env):

Variable	Description
VITE_SERVER_URL	Backend API URL (http://localhost:3000/api)

Run the Application

You need three terminals running simultaneously:

# Terminal 1 — Backend API
cd ai-token-assistant
npm run dev

# Terminal 2 — Inngest Dev Server
cd ai-token-assistant
npm run inngest-dev

# Terminal 3 — Frontend
cd ai-token-frontend
npm run dev

First-Time Setup

1. Open http://localhost:5173 and create an account
2. Promote your user to admin:

   cd ai-token-assistant
   node makeAdmin.js your-email@example.com

3. Log in again — you now have access to the Admin panel

API Reference

Auth Routes

Method	Endpoint	Auth	Description
POST	/api/auth/signup	—	Register a new user
POST	/api/auth/login	—	Login, returns JWT
POST	/api/auth/logout	User	Verify token (stateless)
GET	/api/auth/users	Admin	List all users
POST	/api/auth/update-user	Admin	Update user role & skills

Token Routes

Method	Endpoint	Auth	Description
GET	/api/tokens	User	List tokens (own for users, all for mods/admins)
GET	/api/tokens/:id	User	Get token details
POST	/api/tokens	User	Create token (triggers AI triage)

Data Models

User:

• email — Unique email address
• password — Bcrypt-hashed password
• role — user | moderator | admin
• skills — Array of skill tags (e.g., ["React", "MongoDB"])

Token:

• title — Issue title
• description — Full issue description
• status — TODO | IN_PROGRESS
• priority — low | medium | high (AI-generated)
• helpfulNotes — AI-generated technical notes
• relatedSkills — AI-extracted skill tags
• assignedTo — Assigned moderator/admin reference
• createdBy — Token creator reference

Customization

Component	What to Change
AI Model	Swap Gemini for any LLM in utils/Ai.js
Email Provider	Replace Mailtrap with any SMTP provider in utils/mailer.js
Skills Taxonomy	Update the AI prompt in utils/Ai.js to extract domain-specific skills
Assignment Logic	Modify the matching algorithm in ont-token-create.js
UI Theme	Change DaisyUI theme in Tailwind config

Acknowledgments

• Inngest for the durable workflow engine and agent framework
• Google Gemini for the AI analysis capabilities
• DaisyUI for the UI component library
• Mailtrap for email testing infrastructure

---

Built with autonomous agents. Powered by intelligent automation.