GreenProof

GreenProof is an open-source ESG verification platform that cross-checks corporate sustainability disclosures against real operational data. It combines rule-based NLP claim extraction, real-time IoT sensor comparison, machine learning anomaly detection, and AI-powered analysis to identify inconsistencies in ESG reports automatically.

Built at Geminathon 2026, a 24-hour hackathon organized by the Code Y Gen Club at VIT Chennai, with support from Google Gemini, ElevenLabs, Major League Hacking, and YSOC.

---

Table of Contents

• Overview
• Architecture
• Tech Stack
• Getting Started
• Project Structure
• API Reference
• ML Model
• Roadmap
• Contributing

---

Overview

Most ESG scoring tools take company disclosures at face value. GreenProof verifies them by running a multi-stage pipeline:

1. Claim Extraction — A rule-based NLP parser extracts ESG claims from uploaded PDF reports and classifies them as quantitative, future target, comparative, or general. Mathematical consistency checks are applied where possible.
2. IoT Cross-Verification — Extracted claims are compared against live facility-level sensor data (CO2 emissions, energy consumption, water usage, waste generation) streamed in real time via WebSocket.
3. Anomaly Detection — A custom Isolation Forest model evaluates company metrics against industry norms and flags statistically anomalous patterns that may indicate greenwashing.
4. AI Analysis — Gemini 1.5 Flash performs structured discrepancy analysis between report claims and IoT data, assigning severity levels and confidence scores to each inconsistency.
5. Industry Simulation — A physics-based simulation engine lets users model operational changes and see their precise impact on ESG scores, emissions, and costs before implementation.
6. Suggestion Generation — Gemini generates prioritised improvement recommendations with estimated CO2 reduction, cost savings, and payback periods.
7. Audit Trail — Every verification run is stored in a SHA-256 hash chain, making historical records tamper-evident.

---

Tech Stack

Frontend

• React 19, TypeScript, Vite
• Tailwind CSS v4
• Motion
• Lucide React
• pdfjs-dist
• React Dropzone

Node Backend

• Express.js
• Better-SQLite3
• Node.js crypto module

Python Backend

• FastAPI, Uvicorn
• scikit-learn
• joblib
• pandas, numpy
• google-generativeai
• python-dotenv
• websockets

---

Getting Started

Prerequisites

• Node.js 18 or higher
• Python 3.11 or higher
• A Gemini API key from aistudio.google.com

Installation

1. Clone the repository

git clone https://github.com/YOUR_USERNAME/greenproof.git
cd greenproof

2. Configure environment variables

Create a .env.local file in the project root:

GEMINI_API_KEY=your_gemini_api_key_here

3. Install frontend dependencies

npm install

4. Install Python dependencies

cd backend
pip install -r requirements.txt

5. Train the ML model

Place the ESG dataset at backend/data/esg_data.csv.csv, then run:

cd backend
python train_model.py

This generates backend/models/integrity_model.pkl and backend/models/scaler.pkl. This step is only required once. If the model files are absent, the system falls back to a rule-based integrity checker automatically.

6. Start the Python backend

cd backend
python main.py
# Listening at http://localhost:8000

7. Start the frontend

In a separate terminal:

npm run dev
# Listening at http://localhost:3000

---

Project Structure

greenproof/
├── src/
│   ├── components/
│   │   ├── FileUpload.tsx
│   │   ├── ClaimCard.tsx
│   │   ├── VerificationLog.tsx
│   │   ├── IoTDashboard.tsx
│   │   ├── SimulationControls.tsx
│   │   ├── SuggestionPanel.tsx
│   │   ├── DiscrepancyAnalysis.tsx
│   │   └── IntegrityChecker.tsx
│   ├── services/
│   │   ├── geminiService.ts
│   │   └── pythonApiService.ts
│   ├── lib/
│   │   ├── pdfExtractor.ts
│   │   └── utils.ts
│   ├── types.ts
│   └── App.tsx
├── backend/
│   ├── main.py
│   ├── iot_simulator.py
│   ├── simulator.py
│   ├── ai_analyzer.py
│   ├── suggestion_engine.py
│   ├── ml_integrity.py
│   ├── integrity_engine.py
│   ├── explanation_engine.py
│   ├── train_model.py
│   ├── models/               # Generated .pkl files (gitignored)
│   ├── data/                 # Training data (gitignored)
│   └── requirements.txt
├── server.ts
├── index.html
├── package.json
├── vite.config.ts
└── .env.local                # API keys (gitignored)

---

API Reference

Interactive documentation is available at http://localhost:8000/docs when the backend is running.

Method	Endpoint	Description
GET	/health	Health check and model status
GET	/api/iot/current	Current sensor readings across all facilities
GET	/api/iot/history/{hours}	Historical sensor data
WS	/api/iot/stream	Real-time WebSocket data stream
POST	/api/simulate/scenario	Run an industry simulation scenario
POST	/api/simulate/compare	Compare two simulation scenarios
POST	/api/compare/claims-vs-reality	Gemini discrepancy analysis
POST	/api/suggestions/generate	Gemini improvement suggestions
POST	/api/integrity/check	ML anomaly detection on company data
POST	/api/integrity/check-from-simulation	Simulation piped into ML model
POST	/api/analyze/full	Full end-to-end pipeline in one call

---

ML Model

The integrity checker uses a scikit-learn Isolation Forest trained on the following ESG signals:

Feature	Description
carbon_intensity	CO2 emissions divided by revenue ($M)
energy_intensity	Energy consumption divided by revenue ($M)
water_intensity	Water usage divided by revenue ($M)
ESG_Environmental	Environmental score (0-100)
ESG_Overall	Overall ESG score (0-100)

Configuration: n_estimators=100, contamination=0.07, random_state=42. Features are normalised using StandardScaler prior to training and inference.

At inference the model returns an anomaly score, a binary suspicious flag, and a set of human-readable explanations identifying which intensity ratios triggered the flag.

---

Roadmap

• Vision-based parsing for charts and infographics in PDF reports
• Historical ESG trend tracking per company
• PostgreSQL migration
• User authentication
• Fine-tuning Mistral 7B for ESG claim classification to remove third-party AI dependency
• Real IoT device integration via MQTT
• Multi-company comparison dashboard
• PDF export for verification reports

---

Contributing

Contributions are welcome. Please open an issue before submitting a pull request to discuss the proposed change.

1. Fork the repository
2. Create a feature branch (git checkout -b feature/your-feature)
3. Commit your changes (git commit -m "add your feature")
4. Push to the branch (git push origin feature/your-feature)
5. Open a pull request

---

Acknowledgements

Developed at Geminathon 2026, organized by the Code Y Gen Club at VIT Chennai. Supported by Google Gemini, ElevenLabs, Major League Hacking, and YSOC.