Agentic Campus Assistant

1. What This Project Does

The Agentic Campus Assistant is an AI-powered autonomous conversational system built to simplify everyday university tasks such as booking study rooms, sending official emails, retrieving campus information, and navigating university locations. Through a single voice or chat interface, students can issue natural language commands that the assistant understands, reasons over, and executes end-to-end, eliminating the need to switch between multiple platforms and streamlining campus productivity.

2. Project Workflow

  • Students interact with the assistant using natural text or voice commands such as “Book a study room tomorrow at 3 PM” or “Email my professor about office hours.”
  • Integrated the OpenAI Whisper API to convert voice input into accurate text for hands-free operation.
  • Designed an agentic reasoning architecture powered by Gemini 2.5 Flash that breaks down requests into actionable multi-step plans.
  • Implemented the Model Context Protocol (MCP) to dynamically manage tool selection and route tasks to appropriate APIs or automation modules.
  • Built a Retrieval-Augmented Generation (RAG) pipeline combining the Google Search API for real-time campus data retrieval with ChromaDB for vector-based memory and semantic grounding.
  • Used LangChain to orchestrate prompt chaining, contextual memory, reasoning workflows, and tool execution.
  • Integrated Gmail API to autonomously draft and send emails based on conversational intent.
  • Used Playwright automation to book study rooms and submit university forms through controlled browser workflows.
  • Connected the Google Maps API to provide campus navigation, route generation, and location insights.
  • Added a human-in-the-loop verification layer to ensure user approval for sensitive actions such as sending emails or confirming reservations.
  • Developed a Gradio-based frontend supporting real-time chat and voice conversation for an intuitive user experience.

3. Results and Impact

The Agentic Campus Assistant achieved measurable improvements in student productivity by reducing task completion time by approximately 70%. Users completed complex workflows with minimal input, benefiting from voice-controlled accessibility, intelligent task orchestration, and seamless real-world execution. The system demonstrated true agentic behavior by independently planning, reasoning, and acting to accomplish user goals.

4. Summary

This project combines advanced agentic AI reasoning with real-world automation to create a unified campus productivity platform. By integrating multi-step planning, retrieval-augmented generation, dynamic tool selection via MCP, and direct API/browser automation, the assistant enables students to delegate everyday campus tasks to a single conversational interface, saving time while improving efficiency, accessibility, and engagement.

Tech Stack

  • Gemini 2.5 Flash LLM
  • LangChain
  • Model Context Protocol (MCP)
  • ChromaDB
  • Retrieval-Augmented Generation (RAG)
  • Google Search API
  • Gmail API
  • Google Maps API
  • Playwright Automation
  • OpenAI Whisper (Speech-to-Text)
  • Gradio