PDF RAG System

Installation

1. Install uv (Requires Python 3.10+):

pipx install uv

2. Clone the repository:

git clone <repository-url>
cd advanced-rag

3. Create and activate virtual environment using uv:

uv venv
.venv\Scripts\activate

4. Install dependencies using uv sync (reads from pyproject.toml):

uv pip sync pyproject.toml

5. Install and start Ollama:

• Download from Ollama's website
• Pull the Llama3.2 model:

ollama pull llama3.2

Note: uv sync ensures exact dependency resolution from pyproject.toml, providing faster and more reliable package installation.

Features and Usage

Key Features

Basic RAG

• Simple PDF document loading and parsing
• Basic text chunking
• Vector store using FAISS
• Question-answering with Llama2

Advanced RAG

• Enhanced document chunking with semantic boundaries
• BAAI/bge-large-en-v1.5 embeddings
• Contextual compression
• Source attribution and metadata tracking
• Similarity score filtering
• Custom prompt templates
• Multi-document context handling

Example Usage

1. Place your PDFs in the data folder:

data/
  ├── document1.pdf
  ├── document2.pdf
  └── document3.pdf

2. Run a script for basic pdf based simple RAG. It returns answer to simple query "What is the main topic of the PDF documents?"

 python .\pdf_rag.py
 

3. Run a script for advanced pdf based RAG. It returns answer to query "who is Marcus Aurelius?"

 python .\advanced_pdf_rag.py