A Survey on Retrieval-Augmented Generation for Multi-PDF Question Answering Systems

Retrieval-Augmented Generation (RAG) effectively separates knowledge retrieval from generation for private document question answering. This study explores the design space of document-grounded QA and details a multi-PDF RAG system. The pipeline uses PyMuPDF for text extraction with OCR fallback, BAAI/bge-small-en-v1.5 for dense embeddings, and ChromaDB for storage. Semantically similar chunks are retrieved via cosine distance to generate answers using a local Llama 3.2:3B model via Ollama. Built with React/Vite and FastAPI, the system supports document management and query APIs. The paper reviews RAG methodologies—including hybrid retrieval, reranking, and GraphRAG—while maintaining a local, modular architecture optimized for consumer-grade hardware. Finally, it evaluates the system's strengths and limitations to identify future research directions.