🧠 Feynman AI System

An intelligent learning assistant that applies the Feynman Technique to simplify academic text, verify factual accuracy using NLI, and generate comprehension quizzes.

🏗️ Project Structure

feynman-system/
├── frontend/        # React.js UI
├── backend/         # Node.js + Express API server
└── nli-service/     # Python Flask NLI verification microservice

---

✨ Features

• 📖 Feynman Simplification — Explains complex academic text like you're 10 years old
• 🔍 NLI Fact Verification — Uses DeBERTa v3 to verify simplified claims against the original
• 🔄 Auto-Regeneration — Automatically corrects low-accuracy explanations
• 🧩 Quiz Generation — Creates 4-question MCQ quizzes from the source text
• 🤖 Dual LLM Support — Gemini 2.5 Flash with Groq (LLaMA 3.3 70B) fallback

---

🚀 Getting Started

Prerequisites

• Node.js 18+
• Python 3.9+
• Gemini API key
• Groq API key (optional, for fallback)

---

1. NLI Service (Python)

cd nli-service
pip install -r requirements.txt
python app.py
# Runs on http://localhost:5001

---

2. Backend (Node.js)

cd backend
cp .env.example .env
# Edit .env and add your API keys
npm install
node server.js
# Runs on http://localhost:5000

---

3. Frontend (React)

cd frontend
npm install
npm start
# Runs on http://localhost:3000

---

⚙️ Environment Variables

Create backend/.env from backend/.env.example:

Variable	Description
GEMINI_API_KEY	Google Gemini API key
GROQ_API_KEY	Groq API key (fallback LLM)
HTTP_REFERER	Optional - your app's URL

---

🧪 How It Works

User Input (academic text)
        │
        ▼
   Backend (Node.js :5000)
        │
        ├──► Gemini 2.5 Flash → Simplify text (Feynman style)
        │
        ├──► Extract factual claims from simplified text
        │
        ├──► NLI Service (Python :5001) → Verify claims vs original
        │
        ├──► If score < 80% → Regenerate & re-verify
        │
        └──► Generate MCQ quiz
        │
        ▼
   Frontend (React :3000) → Display results

---

📦 Tech Stack

Layer	Technology
Frontend	React.js, Axios
Backend	Node.js, Express, Axios, dotenv
NLI Service	Python, Flask, HuggingFace Transformers
LLM	Gemini 2.5 Flash + Groq LLaMA 3.3 70B
NLI Model	DeBERTa-v3-small, all-MiniLM-L6-v2