QuickRAG

QuickRAG is a Python library that implements a Retrieval-Augmented Generation (RAG) pipeline for question answering on PDF documents. It combines document processing, embedding generation, and language model inference to provide context-aware answers to user queries.

Features

• PDF processing and text extraction
• Text chunking and embedding generation
• Efficient similarity search for relevant context retrieval
• Integration with Hugging Face Transformers for language model inference
• Support for quantization to optimize memory usage and inference speed

Installation

pip install quickrag

Usage

Here's a basic example of how to use QuickRAG:

from quickrag import QuickRAG

# Initialize QuickRAG
rag = QuickRAG("path/to/your/document.pdf", huggingface_token="YOUR_HUGGINGFACE_TOKEN")

# Process the PDF and create embeddings
rag.process_pdf()
rag.create_embeddings()

# Load the language model
rag.load_llm()

# Ask a question
query = "What are the macronutrients, and what roles do they play in the human body?"
answer = rag.ask(query)

print(f"Query: {query}")
print(f"Answer: {answer}")

Configuration

QuickRAG can be customized with the following parameters:

• pdf_path: Path to the PDF document
• embedding_model_name: Name of the sentence transformer model for embeddings (default: "all-mpnet-base-v2")
• llm_model_name: Name of the language model for answer generation (default: "google/gemma-2b-it")
• use_quantization: Whether to use quantization for the language model (default: True)
• huggingface_token: Your Hugging Face API token

Requirements

• Python 3.7+
• PyTorch
• Transformers
• Sentence-Transformers
• PyMuPDF
• spaCy
• NumPy
• Pandas

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgements

• Hugging Face for their Transformers library
• Sentence-Transformers for the embedding models
• PyMuPDF for PDF processing