# LLM Hallucination Detector Guidelines

## Commands
- Setup: `pip install -r requirements.txt`
- Configure: Set environment variables `HF_MISTRAL_API_KEY` and `HF_OPENAI_API_KEY`
- Run: `python app.py`
- Lint: `ruff check app.py`
- Format: `black app.py`
- Type check: `mypy app.py`

## Code Style
- Follow PEP 8 conventions with 4-space indentation
- Use type hints with Pydantic for data validation
- Write descriptive docstrings using triple quotes
- Name variables/functions in snake_case, classes in PascalCase
- Organize imports: stdlib first, then third-party, then local
- Exception handling: use try/except blocks with specific exceptions
- Constants should be UPPERCASE and defined at class/module level
- Prefer f-strings over other string formatting methods

## Architecture
- App uses Gradio for UI, SQLite for persistence
- LLM integration with Mistral Large and OpenAI o3-mini
- Paraphrase-based approach for hallucination detection
- Maintain clean separation between UI and backend logic