
About the project.

UniHelp AI is a bilingual (Arabic/English) conversational assistant purpose-built for Nahda University — Faculty of Computers and Information. It delivers instant, grounded answers to student queries across 6 knowledge domains: Course Registration, Academic Calendar, Fees & Payments, Results & GPA, Exams & Timetables, and Portal Help. The system constrains the LLM to verified institutional JSON data — eliminating hallucination at the architectural level. The backend is a Django 6 + DRF API deployed on Railway, with a React 18 frontend on Netlify CDN.
Features
>The LLM is constrained exclusively to verified institutional JSON data — refusing to answer anything outside its knowledge base, eliminating hallucination at the architectural level.
Handles mixed-language queries natively, with explicit code-switching detection rules in the system prompt. Students can ask in Arabic, English, or a mix and receive coherent responses.
Covers Course Registration, Academic Calendar, Fees & Payments, Results & GPA, Exams & Timetables, and Portal Help — each backed by structured JSON data loaded per-intent.
Groq Cloud's Llama 3.3 70B delivers sub-second response latency, enabling a genuinely real-time conversational experience at scale.
Designed with a layered architecture — Intent Classifier routes queries to domain-specific handlers, injecting the relevant JSON context into the LLM prompt without any API contract changes.
Client-side XSS sanitization, CORS whitelisting, CSRF protection, Django security middleware, HSTS, and prompt-level guardrails work together across every layer.
Features
>Tech Stack
/>JavaScript library for building user interfaces with reusable components.
A utility-first CSS framework providing pre-designed, atomic-level utility classes for rapid UI development with minimal CSS code.
High-level Python web framework that encourages rapid development and clean, pragmatic design.
A high-level, general-purpose programming language.
Ultra-fast LLM inference API powering Llama 3.3 70B with sub-second latency for real-time AI responses.
Global CDN platform for deploying the React SPA with automatic HTTPS and edge delivery.
Infrastructure platform that enables instant deployments.
Powerful, open source object-relational database system.
