Abstract

CatAssist was developed as a web-based academic planning and advising assistant for School of Information Technology students. The platform integrated a Next.js front end, a Neon-hosted PostgreSQL database, and the Google Gemini 1.5 Flash API to deliver real-time, data-driven answers to student questions. A relational schema was implemented to model course catalogs, faculty directories, degree requirements, and student progress. Retrieval-augmented generation (RAG) was used to ground responses in verified database records and reduce hallucinated output. System testing showed that the assistant correctly synthesized multi-table joins to return class locations, instructor details, prerequisite chains, and remaining requirement summaries from typical queries. The MVP also supported account creation, dashboard visualization, and CSV/ICS event ingestion for weekly schedule generation. Overall, the implementation demonstrated that AI-guided, database-backed advising could improve information accessibility and reduce routine workload for human advisors. Role-based access and environment-managed secrets were applied to protect student data during deployment and testing.