ngirimana/itegeko_ai

Itegeko AI is a production-grade legal information platform designed to make Rwanda's laws accessible through modern technology. It combines high-performance microservices, vector search, and Retrieval-Augmented Generation (RAG) to provide accurate, source-backed answers to legal queries. # ⚖️ Itegeko AI - Rwanda Law Navigator Itegeko AI is a production-grade legal information platform designed to make Rwanda's laws accessible through modern technology. It combines high-performance microservices, vector search, and Retrieval-Augmented Generation (RAG) to provide accurate, source-backed answers to legal queries. --- ## ✨ Key Features - **🔍 Semantic Legal Search**: Find relevant laws using natural language, powered by `pgvector` and deep learning embeddings. - **🤖 RAG-Powered Q&A**: Ask complex legal questions and receive answers cited directly from official legal documents. - **🕷️ Smart Scraper**: Automated ingestion pipeline for official Rwanda Law Reform Commission (RLRC) PDFs. - **🔐 Enterprise Identity**: Robust user management, RBAC, and audit logging via Keycloak. - **📂 Document Management**: Integrated S3/MinIO storage for original legal texts and orders. --- ## 🏗️ Architecture Itegeko AI uses a modular microservices architecture built for scale and reliability. ### System Landscape ```mermaid graph TD Client[Browser/Mobile] --> Frontend[Next.js Frontend] subgraph "Core Services" Frontend --> LegalSvc[Legal Service - Spring Boot] Frontend --> IdentSvc[Identity Service - Spring Boot] end subgraph "Intelligence & Storage" LegalSvc --> AISvc[AI Service - FastAPI] LegalSvc --> LegalDB[(Legal DB - PostgreSQL + pgvector)] LegalSvc --> Minio[(MinIO Storage)] end subgraph "Security & Identity" IdentSvc --> IdentDB[(Identity DB - PostgreSQL)] IdentSvc --> Keycloak{{Keycloak Auth}} end subgraph "Ingestion" Scraper[Law Scraper - Python] --> LegalDB Scraper --> AISvc end ``` ### Data Flow (RAG Pipeline) How a user query is processed and answered using source-backed legal context: ```mermaid sequenceDiagram participant U as User participant L as Legal Service participant A as AI Service participant D as Legal DB U->>L: Ask legal question L->>A: Orchestrate RAG A->>L: Search for context L->>D: Vector Similarity Search D-->>L: Context Snipp …