Logo Lanfrica

fasiryon/liberia-learn

Domain:

educationnatural language processing

Record type:

softwareproject
Creator:
fas
Host:
Modern AI-Driven Education System — dashboards for students & teachers, homework grading, placement testing, AI chat tutor, class management, and real-time progress analytics. Built with Next.js App Router + Prisma + OpenAI. # LiberiaLearn — Production AI Education Platform 🚀 AI-powered platform designed for national-scale deployment (1M+ students, 5,000+ schools) --- ## What This Is LiberiaLearn is a full-stack, production-grade AI system — not a demo — built to deliver real-world education infrastructure. Built and operated by a single engineer (end-to-end system ownership), the platform is designed to function as national education infrastructure. It combines: - RAG-based AI tutoring (grounded responses, not hallucinations) - Multi-tenant SaaS architecture (students, teachers, admins, MOE) - Offline-first learning system for low-connectivity environments - AWS-backed distributed infrastructure - AI-driven workflows for tutoring, grading, and curriculum generation --- ## Live System 👉 liberia-learn.vercel.app --- ## Source Code 👉 github.com --- ## Key Metrics - Tests: 1,577 passing - Test files: 214 - API routes: 189 - Prisma models: 81 - Curriculum: 1,306 READY lessons - Avg lesson length: 1,450 words --- ## Traction - 9,000+ repository interactions in recent weeks - Increasing developer and system-level interest --- ## Technical Highlights - Multi-tenant role system (Student, Teacher, Guardian, Admin, MOE) - Full App Router platform with 189 route handlers - Prisma domain model with 81 models (users, curriculum, AI usage, governance) - RAG-based AI architecture with retrieval + grounding + cost controls - Offline-first system with IndexedDB sync + resumable sessions - Immutable audit logging and governance reporting - SLO tracking, health monitoring, and incident-ready architecture - Full curriculum system with 1306/1306 lessons READY --- ## AI System Design - Retrieval: pgvector cosine similarity over curriculum chunks - Grounding: responses constrained strictly to retrieved context - Providers: OpenAI (primary), Groq (fallback routing) - Prompting: structured templates with role, context, and curriculum injection - Ev …