AI Agent - Multi-Agent RAG System

A sophisticated AI-powered chat application built with Next.js, Mastra framework, and Supabase that implements a multi-agent workflow for intelligent document retrieval and response generation.

🚀 Features

• Multi-Agent Workflow: Orchestrated AI agents for query classification, document retrieval, and response generation
• RAG (Retrieval Augmented Generation): Semantic search through knowledge base using vector embeddings
• Real-time Chat Interface: Modern, responsive chat UI with streaming responses
• Authentication: Secure user authentication with Supabase Auth
• Vector Search: Advanced document similarity matching with configurable thresholds
• Multiple AI Providers: Support for Google Gemini, OpenAI, and Mistral models

📁 Project Structure

├── app/                          # Next.js app directory
│   ├── (auth)/                   # Authentication routes (login, callback)
│   ├── api/                      # API routes and endpoints
│   ├── globals.css               # Global styles
│   ├── layout.tsx                # Root layout component
│   └── page.tsx                  # Home page component
├── components/                   # Reusable React components
│   ├── chat/                     # Chat-related components (interface, message bar)
│   ├── common/                   # Shared UI components (sidebar)
│   └── ProtectedRoute.tsx        # Route protection wrapper
├── hooks/                        # Custom React hooks
│   └── useAuth.ts                # Authentication hook
├── lib/                          # Core library code
│   ├── mastra/                   # Mastra framework configuration
│   │   ├── agents/               # AI agents (classifier, RAG, responder)
│   │   ├── tools/                # Custom tools (RAG tool for document retrieval)
│   │   ├── workflow/             # Multi-agent workflow orchestration
│   │   └── index.ts              # Mastra instance configuration
│   ├── supabase.ts               # Supabase client configuration
│   └── utils.ts                  # Utility functions
├── scripts/                      # Database and testing scripts
│   ├── check-database.ts         # Database connectivity verification
│   ├── debug-embeddings.ts       # Embedding debugging utilities
│   ├── populate-embeddings.ts    # Knowledge base seeding script
│   └── test-rag-tool.ts          # RAG functionality testing
├── services/                     # Business logic services
│   └── conversation.services.ts  # Conversation management
├── supabase/                     # Database schema and migrations
│   ├── complete-setup.sql        # Complete database setup
│   ├── seed.sql                  # Sample data seeding
│   └── *.sql                     # Various SQL scripts
├── .env.local                    # Environment variables
├── package.json                  # Dependencies and scripts
└── README.md                     # Project documentation

🛠️ Tech Stack

• Frontend: Next.js 15, React 19, TypeScript, Tailwind CSS
• Backend: Hono API framework, Vercel Edge Runtime
• AI Framework: Mastra Core with multi-agent workflows
• Database: Supabase (PostgreSQL with vector extensions)
• AI Models: Google Gemini, OpenAI, Mistral
• Authentication: Supabase Auth
• Styling: Tailwind CSS with custom components

📋 Prerequisites

• Node.js 18+ and npm/pnpm
• Supabase account and project
• Google AI API key (for Gemini models)
• Optional: OpenAI or Mistral API keys

🚀 Getting Started

1. Clone and Install

git clone <repository-url>
cd ai-agent
npm install

2. Environment Setup

Create .env.local file:

# Supabase Configuration
NEXT_PUBLIC_SUPABASE_URL=your_supabase_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key
SUPABASE_SERVICE_ROLE_KEY=your_service_role_key

# AI Provider Keys
GOOGLE_GENERATIVE_AI_API_KEY=your_google_ai_key
OPENAI_API_KEY=your_openai_key (optional)
MISTRAL_API_KEY=your_mistral_key (optional)

# Database
DATABASE_URL=your_supabase_database_url

3. Database Setup

Run the database setup scripts:

# Execute the SQL files in your Supabase SQL editor
# 1. supabase/complete-setup.sql
# 2. supabase/seed.sql (optional sample data)

4. Populate Knowledge Base

npm run seed:embeddings

5. Development

# Start Next.js development server
npm run dev

# Start Mastra development server (in another terminal)
npm run dev:mastra

Visit http://localhost:3000 for the application and http://localhost:4111 for Mastra playground.

🔧 Available Scripts

• npm run dev - Start Next.js development server
• npm run build - Build production application
• npm run start - Start production server
• npm run dev:mastra - Start Mastra development server
• npm run build:mastra - Build Mastra workflows
• npm run seed:embeddings - Populate knowledge base with embeddings
• npm run verify:setup - Verify database and API setup
• npm run test:rag - Test RAG functionality
• npm run check:database - Check database connectivity

🤖 How It Works

Multi-Agent Workflow

1. Query Classification Agent: Analyzes user queries to determine intent and category
2. RAG Agent: Retrieves relevant documents from the knowledge base using semantic search
3. Responder Agent: Generates contextual responses based on retrieved information

RAG Implementation

• Documents are chunked and embedded using Google's text-embedding-004 model
• Vector similarity search with configurable thresholds (0.6-0.8)
• Supports category filtering and result count limits
• Real-time document retrieval with similarity scoring

Authentication Flow

• Supabase Auth integration with email/password and social providers
• Protected routes with automatic redirection
• Session management with React hooks

🔒 Security Features

• Row Level Security (RLS) policies in Supabase
• API route protection with authentication middleware
• Environment variable validation
• Secure token handling

📚 Key Components

• Chat Interface: Real-time messaging with streaming responses
• RAG Tool: Semantic document retrieval with vector search
• Multi-Agent System: Orchestrated AI workflow for complex queries
• Knowledge Base: Vector-enabled document storage and retrieval
• Authentication: Secure user management and session handling

🚀 Deployment

The application is optimized for Vercel deployment with Edge Runtime support:

npm run build
# Deploy to Vercel or your preferred platform

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

This project is licensed under the MIT License.