Authors: Meganathan R, Ramkumar R, Indhuja N
Abstract: Personal AI assistants are increasingly used to support information access, learning, programming, and everyday productivity. However, a useful assistant should combine natural-language interaction with accessible controls, conversation context, voice input and output, and a responsive interface. This paper presents ARIYAAN, a full-stack personal AI assistant designed using React, Vite, Node.js, Express, and the Google Gemini API through the official @google/genai SDK. The application supports conversational question answering, programming assistance, mathematical explanations, Markdown-formatted responses, code highlighting, multi-turn conversation context, and locally stored conversation sessions. Browser-native Web Speech APIs provide speech recognition and speech synthesis where supported by the user’s browser. A voice-intent layer handles selected commands, including greetings, time and date requests, assistant identity questions, and desktop application launch requests. A settings interface supports model selection, voice preferences, speech rate, theme selection, and API connection testing. Gemini requests are routed through a backend service so that the API credential does not need to be embedded in frontend source code. The paper describes the system architecture, functional modules, implementation approach, and an evaluation plan covering response functionality, voice interaction, usability, security, and error handling. As measured performance data have not been supplied, this paper does not claim unverified accuracy, latency, or usability results. The proposed application is intended as an educational and productivity-support tool, with future work focused on persistence, accessibility, multilingual interaction, and stronger system-control safeguards.
DOI: https://doi.org/10.5281/zenodo.23560271
