PDFgpt: Chat with your PDFs Using Generative AI

This project introduces a PDF-based Question Answering (QnA) system leveraging Large Language Models (LLMs) to enable efficient and accurate information retrieval from diverse document formats. Using PyMuPDF, PyPDF2, and OCR through PyTesseract, the system extracts and processes both text-based and image-based PDF content. It then utilizes Sentence Transformers for embeddings, stored in Chroma's vector database for high-performance similarity search. This architecture supports LLaMA3-8b-8192 for contextdriven responses delivered via text and audio through gTTS. Future enhancements include multilingual support, multi-PDF querying, and real-time indexing, providing a versatile solution for complex document interaction across domains