Multi-Agent RAG for University Knowledge Management

A question-answering assistant for the circulars, regulations and syllabi of a university, built as a multi-agent retrieval-augmented generation (RAG) system that runs on a student laptop CPU. A router agent sends each question to specialist agents for circulars, regulations and syllabi, each with its own tools; a synthesizer writes the answer with a citation after every fact, and a verifier checks the citations and abstains when the documents do not cover the question. On 265 evaluation questions the multi-agent system answers 55.5 % correctly against 40.8 % for a single-agent hybrid RAG, and abstains on 100 % of the unanswerable questions.

Category: Natural Language Processing / Retrieval-Augmented Generation / Multi-Agent Systems / Knowledge Management

CSE Project 14 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.