akinsola-data/afri-juris-rag

An enterprise Retrieval-Augmented Generation (RAG) microservice deployed via Docker & Streamlit. Instantly answers complex corporate governance queries regarding Central Bank of Nigeria (CBN) regulations, BOFIA 2020, and commercial tax laws with zero hallucination. --- title: AfriJuris RAG emoji: ⚖️ colorFrom: blue colorTo: indigo sdk: docker pinned: false license: mit --- # AfriJuris RAG — AI Legal Assistant for Nigerian Commercial Law & CBN Banking Regulations > **One-line pitch:** A production-grade Retrieval-Augmented Generation (RAG) microservice deployed via Docker & Streamlit that instantly answers complex corporate governance queries regarding Central Bank of Nigeria (CBN) regulations, BOFIA 2020, and commercial tax laws. --- ## Executive Summary Navigating the regulatory environment of Africa's largest economy is a massive challenge for multinational corporations, foreign investors, and fintech scaling across Nigeria. Between dynamic Central Bank of Nigeria (CBN) policy circulars, complex tiered Corporate Income Tax (CIT) structures, and rigorous statutory banking compliance under BOFIA 2020, regulatory breaches carry existential risks—including statutory administrative fines up to ₦50,000,000 or full banking license revocation. A critical vulnerability of generic Large Language Models (LLMs like ChatGPT or Claude) is **hallucination**—they frequently invent fake legal precedents or fail to access unindexed, localized CBN circulars and Nigerian federal tax amendments. **AfriJuris RAG** solves this critical enterprise vulnerability by implementing an advanced Retrieval-Augmented Generation architecture. By anchoring AI responses strictly in an authenticated, localized Vector Database of Nigerian financial law, AfriJuris guarantees absolute legal accuracy with zero hallucination. ### Key Accomplishments: 1. **Comprehensive Regulatory Ingestion:** Built an automated ingestion and chunking engine that processes dense legal texts across BOFIA 2020, CBN Foreign Exchange Manuals, Nigeria Finance Act 2023, and the NDPA 2023. 2. **Local Dense Vector Embeddings:** Engineered an optimized TF-IDF bigram/trigram vector space model (`ngram_range=(1, 3)`) that captures specific statutory syntax such as *"Electronic Money Tr …