NOMOS

A Retrieval-Augmented Generation framework for jurisdiction-specific legal question answering over Indian statutory corpora — presented at ICETCI 2026 (Mahindra University) and under publication with IEEE Xplore.

NOMOS grounds answers in a curated corpus of ten major Indian Acts plus a legal dictionary (~2.32M characters, including the Bharatiya Nyaya Sanhita 2023). Legal-aware chunking (RecursiveCharacterTextSplitter, 2048 / 200 overlap, Section/Article separators) feeds all-MiniLM-L6-v2 embeddings into a ChromaDB vector store; the top-k retrieved passages plus the last five conversational turns are injected into a Llama 3.2 prompt served locally via Ollama. Every response is designed to cite the relevant Act, Article and Section, and inference is entirely local — the paper reports no outbound network calls during testing. React chat frontend, Flask API, Google Sign-In.