🏗 What-If RAG: Explainable Construction Project Intelligence

An AI-powered construction project management platform that provides explainable insights, scenario simulations, and intelligent document analysis through an intuitive chat-based experience.

Built by The Multiverse Task Force — Sparsh Agarwal & Divy Dobariya

▶️ Demo Video

---

🎯 Problem Statement

This project addresses two critical challenges in construction and capital planning:

Challenge 1: What-If Analysis with Explainable AI:

• Existing tools cannot simulate project changes in real-time or show impact instantly
• No transparency on why decisions are suggested, making planning slow and unclear

Challenge 2: NLP-based Contract Analysis:

• Key contract details are buried in documents, causing delays and missed risks
• No smart search or automated compliance alerts — review is manual and error-prone

Construction projects suffer from:

• Scattered Data: Critical information buried across PDFs, contracts, BIM models, and reports
• Manual Review Bottlenecks: Hours spent searching for clauses, deadlines, and compliance requirements
• Opaque Dashboards: No understanding of why metrics are flagged or how scores are calculated
• Slow Decision-Making: Unable to quickly simulate "what-if" scenarios for cost overruns or timeline delays

---

✨ Solution: Unified AI-Powered Platform

Core Capabilities

🔍 RAG-Based Document Intelligence — Natural language queries on project documents with source citations

🎚️ Interactive What-If Simulator — Real-time scenario testing with explainable impact analysis using tunable sliders

📊 Visual Insights Dashboard — Highlights, risks, and document chat with transparent reasoning

🔐 Explainable AI — Every insight includes reasoning chains and audit trails

🤖 Dual AI Backend — Choose between OpenAI GPT-4 (production) or Google Gemini (free tier)

---

🚀 Quick Start

Prerequisites

• Python 3.10+
• Node.js 16+
• API Key: OpenAI OR Google AI

---

🔧 Step 1: Backend Setup

Choose Your AI Provider

Option A: OpenAI Backend (Recommended for production)

# Navigate to backend
cd backend

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Configure environment
cp .env.example .env
# Edit .env and add: OPENAI_API_KEY=sk-your_key_here

# Start server
uvicorn main:app --reload --host 0.0.0.0 --port 8000

Option B: Google Gemini Backend (Free alternative)

# Navigate to Gemini version
cd backend/gemini_version

# Create virtual environment
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Configure environment
cp .env.example .env
# Edit .env and add: GOOGLE_API_KEY=your_gemini_key_here

# Start server
uvicorn main:app --reload --host 0.0.0.0 --port 8000

✅ Backend running at: http://localhost:8000
📚 API Docs: http://localhost:8000/docs

---

🎨 Step 2: Frontend Setup

# Open new terminal and navigate to frontend
cd frontend

# Install dependencies
npm install

# Configure backend connection
cp .env.example .env
# Edit .env and add: VITE_API_BASE_URL=http://localhost:8000

# Start development server
npm run dev

✅ Frontend running at: http://localhost:5173

---

📁 Project Structure

whatif-rag-construction/
│
├── backend/                        # OpenAI-powered backend
│   ├── loader.py                   # Multi-format document ingestion
│   ├── vectorstore.py              # ChromaDB + OpenAI embeddings
│   ├── rag_chain.py                # AI reasoning engine
│   ├── main.py                     # FastAPI server
│   └── requirements.txt
│
├── backend/gemini_version/         # Google Gemini backend (free)
│   ├── loader.py
│   ├── vectorstore.py              # ChromaDB + Gemini embeddings
│   ├── rag_chain.py
│   ├── main.py
│   └── requirements.txt
│
└── frontend/                       # React + TypeScript UI
    ├── src/
    │   ├── components/             # Reusable UI components
    │   ├── pages/                  # Main application views
    │   │   ├── Welcome.tsx
    │   │   ├── Chat.tsx
    │   │   ├── ScenarioTuning.tsx
    │   │   └── VisualInsights.tsx
    │   ├── contexts/               # Global state management
    │   └── App.tsx
    └── package.json

---

🏗 Architecture & Workflow

┌─────────────────┐
│ Upload Documents│  ← PDF, DOCX, BIM, CSV, Images
└────────┬────────┘
         │
         ▼
┌─────────────────┐      ┌──────────────┐
│  Document Loader│─────▶│ Vector Store │  ← ChromaDB
│  (Chunking)     │      │  (Embeddings)│
└─────────────────┘      └──────┬───────┘
                                 │
         ┌───────────────────────┼────────────────────┐
         │                       │                    │
         ▼                       ▼                    ▼
┌─────────────────┐   ┌─────────────────┐   ┌─────────────────┐
│  RAG Chat       │   │  Dashboard AI   │   │  What-If Engine │
│  Q&A + Citations│   │  Health Scoring │   │  Scenario Tuning│
└─────────────────┘   └─────────────────┘   └─────────────────┘
         │                       │                    │
         └───────────────────────┴────────────────────┘
                                 │
                                 ▼
                       ┌─────────────────┐
                       │  React Frontend │
                       │  Chat Interface │
                       └─────────────────┘

---

🛠 Technology Stack

Backend

• AI Models: OpenAI GPT-4/GPT-4o | Google Gemini Pro
• RAG Framework: LangChain 0.3.14
• Vector Database: ChromaDB 0.5.23
• API Framework: FastAPI 0.120.3
• Document Processing: PyPDF, python-docx, ezdxf, IfcOpenShell, Pandas

Frontend

• Framework: React 18+ with TypeScript
• Build Tool: Vite
• Routing: React Router
• State Management: Context API
• Styling: Tailwind CSS (via UI components)

---

🎮 Key Features

1. RAG-Powered Chat Interface

• Ask natural language questions about uploaded documents
• Get instant answers with source citations
• Conversational follow-ups maintain context
• Example: "What is the payment schedule?" → Returns answer with contract section reference

2. Interactive Scenario Tuning

• 5 Adjustable Sliders (scored 1-5):
  • Cost Management
  • Timeline Adherence
  • Design Quality
  • Regulatory Compliance
  • Sustainability
• Real-time Analysis: View strengths, weaknesses, and top 3 recommended changes
• What-If Simulation: See instant impact of parameter adjustments
• Example: "What if I reduce timeline score from 4 to 2?" → Shows impact + mitigation steps

3. Visual Insights Dashboard

• Highlights: Key achievements and positive indicators
• Risks: Identified threats requiring attention
• Document Chat: Ask specific questions with contextual answers
• Example: "Which clauses address force majeure events?" → Extracts and cites relevant sections

4. Explainable AI

• Every response includes reasoning chain
• Source citations for all claims
• Audit trails for compliance and transparency

---

📂 Supported File Formats

Format	Use Case	Processing
PDF	Contracts, reports	Page-based chunking
DOCX	Proposals, change orders	Text extraction
CSV/XLSX	Cost sheets, schedules	Table summarization
IFC/BIM	3D building models	Entity extraction
DWG/DXF	CAD drawings	Layer analysis
Images	Site photos, diagrams	Tesseract OCR

---

🔧 API Endpoints

Endpoint	Method	Description
/upload	POST	Upload and index documents
/search	POST	Ask questions (RAG Q&A)
/dashboard	GET	Get AI project health scores
/highlights	GET	Extract key terms and risks
/ask	POST	Run what-if scenario simulations

Example: Ask Question

curl -X POST "http://localhost:8000/search" \
  -H "Content-Type: application/json" \
  -d '{"question": "What are the liquidated damages?"}'

---

🧪 Example Workflow

1. Upload Project Documents (via UI)

   • Contract PDF
   • Cost estimate XLSX
   • Structural plans PDF

2. View Dashboard → Get overall health score with AI reasoning

3. Extract Insights → Automated highlights of deadlines, risks, compliance gaps

4. Run Scenario → "What if steel delivery is delayed 4 weeks?" → Impact analysis + 3-step mitigation

5. Ask Questions → "What are the payment milestones?" → Answer with section citations

---

🚀 Production Deployment

Backend

# Build for production
pip install -r requirements.txt

# Run with Gunicorn
gunicorn main:app -w 4 -k uvicorn.workers.UvicornWorker --bind 0.0.0.0:8000

Frontend

# Build production bundle
npm run build

# Preview locally
npm run preview

# Deploy dist/ folder to hosting service (Vercel, Netlify, AWS S3)

---

🐛 Troubleshooting

Backend won't start:

• Verify Python 3.10+: python --version
• Check API key in .env
• Ensure dependencies: pip install -r requirements.txt

Frontend can't connect:

• Verify backend running: curl http://localhost:8000/about
• Check VITE_API_BASE_URL in frontend .env
• Ensure CORS enabled in backend main.py

Vector search fails:

• Clear ChromaDB: rm -rf /tmp/chroma_store
• Check write permissions

---

🤝 Contributing

We welcome contributions! Please:

1. Fork the repository
2. Create feature branch: git checkout -b feature/AmazingFeature
3. Commit changes: git commit -m 'Add AmazingFeature'
4. Push to branch: git push origin feature/AmazingFeature
5. Open Pull Request

---

📄 License

This project is licensed under the MIT License - see LICENSE file for details.

---

👥 Team

The Multiverse Task Force

• Sparsh Agarwal — Front End and System Integration
• Divy Dobariya — Backend Development & AI/ML, Langchain, RAG Architechture

Built for Aurigo Hackathon 2025 🚀

---

📞 Support

• Issues: GitHub Issues
• Email: divydobariya11@gmail.com

---

⚡ Make better construction decisions, faster, with transparent AI reasoning.

⭐ Star this repo if you find it useful!