---
title: RAG Python System
emoji: 🤖
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 4.44.1
app_file: app.py
pinned: false
license: mit
---

# RAG Python System 🤖

A local-first Retrieval-Augmented Generation (RAG) system that enables intelligent question-answering over your documents using Hugging Face Inference API, ChromaDB, and Gradio.

## Features

- 📄 **Document Processing**: Converts PDF, DOCX, and TXT files to searchable format
- 🔍 **Semantic Search**: Uses sentence transformers for accurate context retrieval
- 🤖 **AI-Powered Answers**: Leverages Llama 3.2 3B via Hugging Face Inference API
- 💬 **Interactive UI**: Clean Gradio interface for easy interaction
- 🎯 **Source Citations**: Provides references to source documents

## How It Works

1. **Document Ingestion**: Upload or use pre-loaded documents (currently includes "Think Python" guide)
2. **Semantic Chunking**: Documents are split into meaningful chunks
3. **Vector Embeddings**: Text chunks are converted to embeddings using `all-MiniLM-L6-v2`
4. **Context Retrieval**: Relevant chunks are retrieved based on your question
5. **Answer Generation**: Llama 3.2 generates answers using the retrieved context

## Setup Instructions

### For Hugging Face Spaces Deployment

1. **Fork or Duplicate this Space**
   - Click the three dots menu → "Duplicate Space"

2. **Get Your Hugging Face Token**
   - Go to [https://huggingface.co/settings/tokens](https://huggingface.co/settings/tokens)
   - Create a new token with `read` permissions
   - Copy the token (starts with `hf_`)

3. **Add Token to Space Secrets**
   - Go to your Space Settings
   - Navigate to "Repository secrets"
   - Add a new secret:
     - **Name**: `HF_TOKEN`
     - **Value**: Your token from step 2
   - Click "Add"

4. **Restart the Space**
   - The Space will automatically rebuild and start

### For Local Development

1. **Clone the Repository**
   ```bash
   git clone https://github.com/monsara/rag-python-rag.git
   cd rag-python-rag
   ```

2. **Create Virtual Environment**
   ```bash
   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   ```

3. **Install Dependencies**
   ```bash
   pip install -r requirements.txt
   ```

4. **Set Environment Variable**
   ```bash
   export HF_TOKEN=hf_your_token_here
   ```

5. **Run the Application**
   ```bash
   python app.py
   ```

6. **Access the Interface**
   - Open your browser to `http://localhost:7860`

## Usage Examples

Try asking questions like:

- "How do if-else statements work in Python?"
- "What are the different types of loops in Python?"
- "How do you handle errors in Python?"
- "Explain Python functions with examples"
- "What is object-oriented programming in Python?"

## Architecture

```
User Query → Gradio UI → Vector Store (ChromaDB) → Context Retrieval
                                                          ↓
                                                   HF Inference API
                                                          ↓
                                                   Llama 3.2 3B
                                                          ↓
                                                   Formatted Response
```

## Tech Stack

- **Frontend**: Gradio 4.44.1
- **LLM**: Llama 3.2 3B Instruct (via Hugging Face Inference API)
- **Embeddings**: all-MiniLM-L6-v2 (Sentence Transformers)
- **Vector DB**: ChromaDB
- **Document Processing**: PyMuPDF, python-docx
- **Text Splitting**: LangChain Text Splitters

## Rate Limits

**Free Tier (Hugging Face Inference API):**
- ~1000 requests/hour
- 1024 max tokens per response
- Shared infrastructure

**For Production:**
Consider upgrading to [Hugging Face Pro](https://huggingface.co/pricing) ($9/month) for:
- Higher rate limits
- Faster inference
- Priority support

## Limitations

- Currently uses a single pre-loaded document ("Think Python")
- Free tier has rate limits
- Response quality depends on context relevance
- Max 1024 tokens per response

## Roadmap

- [ ] File upload functionality
- [ ] Multiple document support
- [ ] Conversation history
- [ ] Custom model selection
- [ ] Advanced filtering options
- [ ] Export conversation feature

## Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

## License

MIT License - see LICENSE file for details

## Links

- **GitHub Repository**: [monsara/rag-python-rag](https://github.com/monsara/rag-python-rag)
- **Documentation**: See [README.md](README.md) for detailed technical documentation
- **Hugging Face**: [Get your API token](https://huggingface.co/settings/tokens)

## Support

If you encounter issues:

1. Check that `HF_TOKEN` is set correctly in Space secrets
2. Verify your token has `read` permissions
3. Check the Space logs for error messages
4. Open an issue on GitHub

---

Built with ❤️ using Hugging Face, Gradio, and ChromaDB