Design and Development of a Full-Stack AI Chat and Image Generation Platform (QuickGPT)

Abstract

This report documents the design and development of QuickGPT, a full-stack AI-powered chat and image generation platform built during an internship at the National Informatics Centre (NIC), Kendrapara. The platform integrates Google Gemini Flash as its core large language model, accessed via the OpenAI-compatible Node.js SDK, and uses ImageKit for AI image generation. The frontend is built with React 19, Vite, and Tailwind CSS, while the backend uses Express.js with MongoDB Atlas for data persistence and JWT-based authentication. A key innovation of the project is a custom real-time data pipeline that classifies user queries for live data requirements and fetches current information from external APIs—including Open-Meteo (weather), CoinGecko (cryptocurrency), Alpha Vantage (stocks), Wikipedia (factual and news), and CricketAPI (sports)—before injecting it into the Gemini prompt context. This effectively implements Retrieval-Augmented Generation (RAG) without a vector database. A Stripe-based credit and subscription system was designed and tested, though temporarily disabled in the live deployment. Both the React frontend and the Express backend are deployed independently on Vercel’s serverless platform. The report covers the full system architecture, database design, backend and frontend implementation, real-time data pipeline, deployment strategy, testing, results, and future work.