README.md

---
title: Jack Patel AI Assistant
emoji: 🦙
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.34.1
app_file: main.py
license: mit
---

# 🤖 Jack Patel AI Assistant

A personalized AI assistant powered by a fine-tuned TinyLlama model, trained to answer questions about Jack Patel using custom data stored in `Data.json`.

## 🌟 Features

- **Personalized Responses**: Trained on 150+ question-answer pairs about Jack Patel (from `Data.json`)
- **Smart Fallback**: If a match isn't found in the training data, the model generates a response using TinyLlama
- **Modern UI**: FastAPI-powered UI using a clean and responsive HTML form
- **API Access**: RESTful API endpoints for programmatic access
- **Container-Ready**: Deployable instantly on Hugging Face Spaces via Docker or source files

## 🚀 How to Use

### 🖥️ Web Interface
Visit the Space and type any question — get a smart response about Jack Patel.

### 🔌 API Usage
Make a GET request to:

```
/api/generate?instruction=your_question
```

### ✅ Health Check
Use:

```
/health
```
to confirm the app is running correctly.

## 📝 Example Questions

- "What is your name?"
- "What is your father's name?"
- "Which school did you attend?"
- "What are your technical skills?"
- "Tell me about your hobbies"

## 🛠️ Technical Details

- **Base Model**: `TinyLlama/TinyLlama-1.1B-Chat-v1.0`
- **Fine-Tuning**: LoRA with your `Data.json` file containing personal Q&A pairs
- **Framework**: FastAPI (backend), Jinja2 (templates)
- **Interface**: HTML/CSS in `templates/index.html`
- **Model Inference**: PyTorch + Transformers

## 📁 File Structure

```
├── main.py                 # FastAPI application logic
├── requirements.txt        # Dependencies
├── Dockerfile             # Build config for Hugging Face Space
├── Data.json              # Personalized Q&A training data
├── templates/
│   └── index.html         # Web-based UI
├── lora_model/            # LoRA fine-tuned TinyLlama model
└── static/                # Optional static files (CSS, JS)
```

## 📊 Model Behavior

- **Uses Data.json**: Loads Q&A pairs at runtime for exact matching
- **Model-Generated Responses**: For non-exact matches, the TinyLlama model generates contextual answers
- **Fallback Intelligence**: Matches similar or rephrased queries using embedding similarity

## 🔧 Setup Instructions

### 🧪 Local Development
```bash
pip install -r requirements.txt
python main.py
```

### ☁️ Hugging Face Spaces
1. Upload all project files including:
   * `main.py`
   * `Data.json`
   * `lora_model/` folder
   * `requirements.txt`
   * `templates/index.html`
2. Hugging Face will automatically detect and run your app.

## 🔒 Privacy & Security

* No data is stored or logged
* Responses are generated locally in-memory
* No external API calls
* Entirely contained in the Hugging Face Space runtime

## 🤝 Contributing

Want to improve this AI assistant? You can:
* Add new question-answer pairs to `Data.json`
* Enhance the UI
* Optimize performance and latency
* Add advanced features (like semantic search)

## 📄 License

This project is licensed under the MIT License.

**Built with ❤️ using FastAPI, PyTorch, and Hugging Face Transformers**