Wisconsin Law Enforcement Legal Chat RAG System

A proof-of-concept Retrieval-Augmented Generation (RAG) system that enables Wisconsin law enforcement officers to quickly query state statutes, case law, and department policies through a conversational interface.

Project Structure

codefourrag/
├── backend/          # FastAPI backend application
├── frontend/         # Next.js frontend application
├── data/            # Documents and embeddings
├── docs/            # Documentation
└── scripts/         # Utility scripts

Setup Instructions

Prerequisites

• Python 3.10+
• Node.js 18+ (for frontend)
• npm or yarn

Backend Setup

1. Create a virtual environment:

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

3. Copy environment variables:

cp .env.example .env
# Edit .env and add your API keys

4. Run the backend:

cd backend
uvicorn main:app --reload

The API will be available at http://localhost:8000

Frontend Setup

1. Navigate to frontend directory:

cd frontend

2. Install dependencies:

npm install

3. Run the development server:

npm run dev

The frontend will be available at http://localhost:3000

Development Workflow

This project is being built incrementally through 10 separate processes. Each process is implemented and tested independently before moving to the next.

API Documentation

Once the backend is running, visit:

• API Docs: http://localhost:8000/docs
• Alternative Docs: http://localhost:8000/redoc

Testing

Run backend tests:

cd backend
pytest

Data Directory

Place your Wisconsin legal documents in the following directories:

• data/raw/statutes/ - State statutes (PDF/HTML)
• data/raw/case_law/ - Case law summaries (PDF)
• data/raw/policies/ - Department policies (DOCX/PDF)
• data/raw/training/ - Training materials

You can organize files within subdirectories as needed. The system will recursively scan all subdirectories within data/raw/.

Supported File Formats

• PDF (.pdf) - Uses pdfplumber
• Word Documents (.docx, .doc) - Uses python-docx
• HTML (.html, .htm) - Uses BeautifulSoup4
• Text Files (.txt, .md) - Plain text parsing

Document Ingestion

Using the API Endpoint

Once the backend is running, you can ingest documents by calling the /api/ingest endpoint:

# Ingest all documents from data/raw/
curl -X POST "http://localhost:8000/api/ingest"

# Or use the interactive API docs at http://localhost:8000/docs

The ingestion process will:

1. Recursively scan data/raw/ and all subdirectories
2. Parse supported file formats (PDF, DOCX, HTML, TXT)
3. Normalize text (remove headers/footers, preserve section markers)
4. Extract metadata (title, jurisdiction, dates, statute numbers, department)
5. Return a list of normalized Document objects

Note: This step does NOT chunk or index documents yet. That will be handled in subsequent steps.

Response Format

The /api/ingest endpoint returns:

• status: "success", "partial", or "failed"
• documents_processed: Number of successfully processed documents
• documents_failed: Number of documents that failed to process
• total_documents: Total number of documents found
• documents: List of Document objects with text and metadata
• failures: List of failed files with error messages
• processing_time_seconds: Time taken to process

Example

{
  "status": "success",
  "documents_processed": 5,
  "documents_failed": 0,
  "total_documents": 5,
  "documents": [
    {
      "text": "Normalized document text...",
      "metadata": {
        "title": "Wisconsin Statute 940.01",
        "jurisdiction": "WI",
        "document_type": "statute",
        "statute_numbers": ["940.01"],
        "dates": ["2023"],
        "source_path": "data/raw/statutes/940.01.pdf"
      },
      "source_path": "data/raw/statutes/940.01.pdf"
    }
  ],
  "failures": [],
  "processing_time_seconds": 2.34
}

Performance Evaluation

To evaluate system performance (retrieval accuracy, response time, relevance scoring):

# Make sure backend is running first
python scripts/evaluate_performance.py

This will generate performance metrics and save results to performance_results.json.

See PERFORMANCE_METRICS.md for detailed methodology and expected results.

Documentation

• README.md: Quick start guide and setup instructions
• EXPLANATION.md: Complete implementation details and technical documentation
• ARCHITECTURE.md: System architecture, design decisions, scalability, and security
• PERFORMANCE_METRICS.md: Performance evaluation methodology and metrics

License

This is a take-home assignment project.