hizkyas/hagersearch-legal-RAG

Hager-Search is a professional, cross-lingual RAG system for Ethiopian legal documents. It provides a conversational interface for the Federal Constitution, the Commercial Code (2021), and strategic frameworks like Digital Ethiopia 2030, delivering precise English legal answers with automated article-level citations # 🇪🇹 Hager-Search: Ethiopian Legal Research RAG Portal **Hager-Search** is a professional, cross-lingual Retrieval-Augmented Generation (RAG) system designed to provide precise, context-aware answers to queries concerning Ethiopian Law. Built with an institutional-grade interface and powered by the latest LLM architectures, Hager-Search transforms complex legal PDF repositories into an interactive, conversational intelligence platform. --- ## 🏛️ Core Mission To democratize access to Ethiopian legal information by bridging the gap between raw PDF proclamations and actionable legal insights through high-performance semantic search and LLM synthesis. ## ✨ Key Features - **Conversational Legal Memory**: Maintains full conversation context using **Llama-3.3 (70B)**, allowing for complex follow-up questions (e.g., *"When was it adopted?"*). - **Automated Source Citations**: Every answer is backed by verifiable citations, automatically extracting **Proclamation No.** and **Article No.** from legal document metadata. - **Cross-Lingual Intelligence**: Accepts queries in both **English and Amharic**, providing precise English legal explanations and citations. - **Professional Legal Dashboard**: A "Split-View" interface featuring a persistent **Legal Repository Sidebar** to show active sources (Constitution, Commercial Code, etc.). - **High-Performance Ingestion**: Uses a robust **PyMuPDF-based pipeline** with Amharic normalization to handle complex, multilingual legal texts efficiently. ## 🛠️ Technical Stack - **AI Engine**: LangChain, Groq (Llama-3.3-70B), Sentence-Transformers (Multilingual MiniLM). - **Vector Database**: ChromaDB (with persistent local storage). - **Backend**: FastAPI (Python 3.11+) with lazy-loading singleton architectures for high performance. - **Frontend**: React.js (Vite) utilizing an authoritative, institutional CSS theme (Oxford Navy & Slate). ## 📊 Project Impact Hager-Search is optimized for legal practitioners, researchers, and citizens, pr …