An adaptive, feedback-based AI tutor system built using:
- 🧠 Meta's LLaMA-3.2-3B-Instruct
- 🔄 LangGraph for multi-agent workflow
- ⚡ Hugging Face Transformers (4-bit quantization for efficiency)
- ✅ PyTorch, BitsandBytes, Accelerate for seamless GPU usage
This notebook walks you through a complete interactive tutor session that:
- 📚 Asks a question from a topic you choose
- 📝 Evaluates your answer and gives structured feedback
- 🧪 Generates a new practice question
- 📈 Tracks your progress and adapts difficulty
It's like having your own AI teacher, personalized to your learning!
You can explore the full .ipynb notebook on Google Colab using the button above.
├── EnhancedTutorSystem.ipynb
├── README.md
├── requirements.txt
This project uses (but does not rehost) Meta's official instruction-tuned model:
The model is loaded via transformers using 4-bit quantization (BitsAndBytes)
Note: You must agree to Meta's license to access the model.
- ✍️ Adaptive questions across difficulty levels
- 📊 Real-time performance tracking
- 🤓 Intelligent feedback on every answer
- 💡 LangGraph-powered multi-agent workflow
- 🧵 Fully reproducible session history
- 🌐 A Hugging Face Space with a user-friendly UI
- 📝 Student progress export to PDF
- 🎯 Topic-based quiz sessions
- 🧪 Integration with LangChain for evaluation metrics
This project is released under the MIT License.
- 🧠 Meta AI for LLaMA-3
- 🔄 LangGraph by LangChain
- 🤗 Hugging Face for open infrastructure
Feel free to raise issues or suggestions on GitHub
Or connect via Hugging Face community tab!
Happy learning! 💡