Authors: Athul Krishna S, Chandrika K, Sanjay G, Sanjay S
Abstract: Students and faculty at engineering colleges often search several portals, notices and documents to find routine academic information. This paper presents Campus AI, a web-based chatbot prototype that lets users ask college-related questions in natural language. The system is implemented as a Next.js and React application. A server-side chat endpoint applies per-IP sliding-window rate limiting and input sanitization, assembles a prompt from an institutional system prompt, the latest eight conversation messages and, when available, the authenticated student’s profile, and then queries the Llama 3.1 70B model hosted on the Groq platform in JSON mode. The reply is returned as a structured object containing the answer text, an optional generated visual and source labels, and is rendered in a conversational interface alongside role-specific dashboards for students, faculty and administrators. We describe the architecture, request workflow and implementation, and we separate the features that are implemented in the prototype from those that are only planned. Functional behaviour is demonstrated through screenshots of the working prototype. No quantitative evaluation of answer accuracy, response time or user satisfaction has been carried out; such an evaluation is identified as future work.
DOI: https://doi.org/10.5281/zenodo.23560791
