A Retrieval-Augmented Generation (RAG) system combines information retrieval with language generation. It works by searching a collection of documents to find relevant content for a user's question, then uses that information to generate a helpful answer.

In this project, the RAG system reads a PDF document, retrieves the most relevant sections based on your query, and presents the answer in a user-friendly web app.

See deployed app version here: https://compliance-rag.streamlit.app/

I used the Cursor agent on Auto model select to help build this project.

Benefits of automating this process include:

• Faster information retrieval: Instantly find relevant content without manual searching.
• Improved accuracy: Reduces the risk of missing important details in large documents.
• Consistency: Ensures the same process is followed every time, minimizing human error.
• Reduced Hallucination: The RAG system only returns information that actually exists in your documents, minimizing the risk of generating made-up or inaccurate answers. This means you get reliable, document-grounded responses instead of AI "hallucinations."
• Time savings: Frees up time for more valuable tasks by automating repetitive lookups.
• Scalability: Easily handles multiple or large documents as your needs grow.
• Checklist Tracking: The system can help maintain a checklist of compliance tasks or requirements, marking off what has already been completed. This ensures nothing is overlooked and provides a clear record of progress.

• Completely Free - No API costs or cloud services required
• Local Processing - Everything runs on your machine
• Simple Setup - Minimal dependencies and configuration
• Persistent Storage - Documents are saved locally

• Python 3.8+
• pip package manager
• virtual environment

1. Install dependencies:

   pip install -r requirements.txt

2. Create sample pdf:

   python scripts/create_sample_pdf.py

3. Run unit tests (optional):

   python -m unittest tests.test_create_sample_pdf -v

4. Run the demo app:

   streamlit run rag_system/web_app.py

rag-tutorial/
├── rag_system/
│   ├── core.py                # RAG core logic
│   ├── utils.py               # RAG utilities
│   └── web_app.py             # Streamlit web app
├── scripts/
│   └── create_sample_pdf.py   # Script to generate sample PDF for testing
├── tests/
│   └── test_create_sample_pdf.py  # Unit tests for PDF creation
├── data/
│   └── documents/             # Folder for PDF documents
├── requirements.txt           # Python dependencies
├── README.md                  # This file
└── ARCHITECTURE.md            # Project architecture

The project includes unit tests to verify that the PDF creation functionality works correctly:

• Test PDF creation: Verifies that the PDF file is generated successfully
• Test content verification: Ensures that expected content appears in the generated PDF
• Test multiple content items: Validates that various sections are included
• Test page count: Confirms the PDF has the expected number of pages

Run all tests with:

python -m unittest tests.test_create_sample_pdf -v

• ChromaDB Documentation
• Hugging Face Transformers
• Sentence Transformers