Secure Offline RAG System
## Secure Offline RAG System
Oct 2024 – Dec 2024
Associated with Veermata Jijabai Technological Institute (VJTI)
This project implements a fully local, offline Retrieval-Augmented Generation (RAG) system optimized for low memory usage and fast CPU-based inference. Built using open-source LLMs, the system enables efficient document querying and question answering over PDFs and other files without relying on cloud services. It combines FAISS-based vector search with an optional reranking model to ensure high-quality, context-aware responses. Featuring multilingual support, dynamic document addition without full reindexing, and an intuitive Streamlit UI, the solution delivers accurate, low-latency retrieval and generation while maintaining a minimal hardware and memory footprint.
Skills: Python, LLMs, FAISS, Streamlit