A decision-support tool that gives compliance and product teams a first-pass, citation-backed assessment of whether an AI use case may trigger EU AI Act concerns. Upload product docs (or describe a use case manually), and Norrin retrieves evidence from your uploads plus a built-in legal corpus (EU AI Act + official guidance). A multi-agent pipeline — Assessment → citation validation → Critic (with optional one revision) → Presenter — produces a structured report with risk tier, governance observations, missing-info questions, and support-labelled citations (strong / moderate / weak / unsupported). Built with Streamlit, Chroma RAG, and DeepSeek. Includes six demo cases (HR screening, chatbot, emotion detection, spam filter, GPAI deployer, predictive maintenance) and a visible agent trace for transparency.
@lmtriet145
owner
