Construction Sustainability Advisor

An AI-powered RAG (Retrieval-Augmented Generation) chatbot that helps construction professionals make informed, sustainable product decisions using Environmental Product Declarations (EPDs) and technical documentation.

🎯 Project Overview

This application demonstrates a production-ready RAG system designed for the construction industry, specifically focused on sustainability metrics and product comparison. It showcases:

• Retrieval-Augmented Generation (RAG) architecture
• Real-time streaming AI responses
• Semantic search through construction product documentation
• Source citation for transparency and verification
• Mock database with realistic EPD data for demonstration

🏗️ Use Cases

• Compare embodied carbon across different material categories
• Find products that meet specific sustainability certifications (BREEAM, LEED, Passive House)
• Identify suitable materials for low-carbon construction
• Verify technical specifications from EPD documentation
• Make data-driven product selection decisions

🚀 Getting Started

Prerequisites

• Node.js 18+ and npm
• (Optional) Gemini API key from Google AI Studio for real AI responses

Installation

# Install dependencies
npm install

# Copy environment template
cp .env.example .env

# (Optional) Add your Gemini API key to .env
# Note: The app works with mock AI responses by default, see examples below.

Running the App

# Start development server
npm run dev

# Open browser to http://localhost:4000

The application will automatically use intelligent mock AI responses if no API key is provided, making it perfect for demonstrations and testing.

🎨 Features

1. RAG-Based Product Search

• Keyword-based retrieval from construction product database
• Context-aware responses using relevant document chunks
• Streaming responses for better UX

2. Mock Data System

• 6 realistic EPD documents covering:
  • Insulation (Rockwool, Kingspan)
  • Concrete (Holcim ECOPact, CemEx Vertua)
  • Timber (Stora Enso CLT)
  • Cladding (Kebony)
• 15 document chunks with technical specifications
• Realistic embodied carbon data and certifications

3. AI Response System

• Primary: Gemini 2.0 Flash API for production use
• Fallback: Intelligent mock responses for demos
• Automatic fallback on API quota/errors
• HTML-formatted tables for data comparison

4. Professional UI

• Clean, modern interface inspired by professional tools
• Responsive design (mobile & desktop)
• Real-time streaming message display
• Source citations with document references
• Example queries for quick testing

🏛️ Architecture

┌─────────────┐
│   Frontend  │ React + React Router
│   (UI/UX)   │
└──────┬──────┘
       │
┌──────▼──────┐
│  API Layer  │ Node.js + Hono
│  (Routes)   │
└──────┬──────┘
       │
   ┌───┴───┐
   │       │
┌──▼───┐ ┌─▼────────┐
│ Mock │ │  Gemini  │
│  DB  │ │ API/Mock │
└──────┘ └──────────┘

Tech Stack

• Frontend: React 18, React Router 7, Tailwind CSS
• Backend: Node.js, Hono framework
• AI: Gemini 2.0 Flash (with smart mock fallback)
• Database: In-memory mock (easily swappable for PostgreSQL/Neon)

📊 Sample Queries

Try these example queries to see the system in action:

1. "What insulation materials have embodied carbon below 2 kg CO2e/kg?"
2. "Compare three concrete products on sustainability metrics"
3. "What products are suitable for passive house construction?"
4. "Show me timber cladding options with EPD certifications"

🔧 Configuration

Using Real Gemini API

1. Get API key from Google AI Studio
2. Add to .env:

   GEMINI_API_KEY=your_api_key_here
   

3. Remove USE_MOCK_AI flag if present

Using Mock AI (Default)

The application automatically uses mock responses when:

• No GEMINI_API_KEY is set
• API quota is exceeded
• Network errors occur

This ensures the demo always works reliably.

📁 Project Structure

construction-product-chatbot/
├── src/
│   ├── app/
│   │   ├── page.jsx              # Main chat interface
│   │   ├── about/page.jsx        # About page explaining RAG
│   │   └── api/
│   │       ├── chat/route.js     # Chat endpoint with RAG logic
│   │       ├── documents/route.js # Documents API
│   │       ├── messages/route.js  # Message history API
│   │       └── utils/
│   │           ├── mockDatabase.js # In-memory product data
│   │           ├── mockAI.js      # Intelligent fallback responses
│   │           ├── gemini.js      # Gemini API integration
│   │           └── sql.js         # Database abstraction
│   └── utils/
│       └── useHandleStreamResponse.js # Stream handling hook
├── .env.example
├── package.json
└── README.md

🚦 Future Enhancements

• Vector embeddings for semantic search
• Real PostgreSQL/Neon database integration
• PDF upload and parsing
• Multi-project conversation management
• Export comparison reports
• Integration with BIM software (Revit, IFC)

📝 License

This project is a demonstration for job application purposes.

👤 Author

Daniele Tatasciore

• LinkedIn
• Portfolio
• GitHub

---

Built with ❤️ using React, Gemini AI, and modern web technologies