🇨🇦 Canadian Criminal Code Assistant

An AI-powered legal assistant for Canadian criminal law, built with Streamlit, Google Gemini AI, and FAISS vector search.

✨ Features

• Intelligent Search: Semantic search through the entire Criminal Code of Canada
• AI-Powered Responses: Context-aware answers using Google Gemini AI
• Source Citations: Always shows relevant legal sections with full references
• Custom Context: Upload legal documents for specialized searches
• Chat History: Track previous questions and responses
• Downloadable Reports: Export questions, answers, and sources
• Professional UI: Clean, responsive interface with legal disclaimers

🚀 Quick Start

1. Install Dependencies

pip install -r requirements.txt

2. Environment Set Up

Create a .env file in the root directory:

GOOGLE_API_KEY=your_gemini_api_key_here

3. Data Preparation

Criminal Code PDF From (https://laws-lois.justice.gc.ca/eng/acts/C-46/page-1.html):

data/
└── C-46.pdf  # Canadian Criminal Code PDF

4. Setup Pipeline

Step 1: Extract Sections from PDF

python pdf_scraper.py

This will:

• Extract legal sections from the PDF
• Clean and validate content
• Save to data/processed/sections.json

Step 2: Build Search Index

python build_index.py

This will:

• Create vector embeddings for all sections
• Build FAISS search index
• Save to faiss_index/ directory

Step 3: Launch the App

streamlit run app.py

📁 Project Structure

.
├── app.py                # Main Streamlit application
├── pdf_scraper.py        # PDF extraction and processing
├── build_index.py        # FAISS index creation
├── requirements.txt      # Python dependencies
├── .env                  # Environment variables (LLM API Key)
├── data/
│   ├── C-46.pdf          # Criminal Code PDF (you provide)
│   └── processed/
│       └── sections.json # Extracted sections (generated)
└── faiss_index/          # Vector search index (generated)

🔧 Advanced Configuration

Embedding Model

The default embedding model is all-MiniLM-L6-v2. To use a different model, modify build_index.py:

self.embedding_model = "sentence-transformers/all-mpnet-base-v2"  # Higher quality

Search Parameters

Adjust search results in app.py:

relevant_docs = assistant.get_relevant_documents(query, k=10)  # More results

Gemini Model

Switch to a different Gemini model in app.py:

model = genai.GenerativeModel(model_name="models/gemini-1.5-pro-001")  # More powerful

🐛 Troubleshooting

Common Issues

1. "FAISS index not found"

   • Run python build_index.py to create the search index

2. "Sections file not found"

   • Run python pdf_scraper.py to extract sections from PDF

3. "GOOGLE_API_KEY not found"

   • Add Gemini API key to the .env file

4. Memory issues during indexing (Optional)

   • Reduce batch size in build_index.py
   • Use a smaller embedding model

5. PDF extraction errors

   • Ensure data/C-46.pdf exists and is readable
   • Check PDF format compatibility

Performance Tips

• First run: Initial model loading takes 30-60 seconds
• Subsequent runs: Models are cached for instant responses
• Large PDFs: Extraction may take several minutes
• Memory usage: ~2GB RAM recommended for full functionality

📋 Usage Examples

Basic Questions

• "What constitutes assault under Canadian law?"
• "What are the penalties for theft over $5000?"
• "How is murder defined in the Criminal Code?"

Section-Specific Queries

• "Explain section 265 of the Criminal Code"
• "What does section 322 say about theft?"

Procedural Questions

• "What is the difference between summary and indictable offences?"
• "How does the bail system work in Canada?"

⚖️ Legal Disclaimer

This application provides general information about Canadian criminal law for educational purposes only. It is NOT a substitute for professional legal advice. Always consult with a qualified lawyer for specific legal matters.

🤝 Contributing

Contributions are welcome! Please ensure all changes maintain:

• Code quality and documentation
• Legal accuracy and appropriate disclaimers
• Performance optimizations
• User experience improvements

📄 License

This project is for educational and research purposes. Please respect copyright laws and use responsibly.

---

Built with: Streamlit • Google Gemini AI • FAISS • LangChain • PyMuPDF