An AI-powered chatbot designed to teach university students how to promote Equity, Diversity, and Inclusion (EDI) within collaborative teams.
Live Demo: poly-edi-chatbot.vercel.app
Case Study: https://www.builtbyseb.dev/projects/edi-ai-chatbot
In late 2025, for a humanities class focused on team collaboration, my team was challenged to teach the class about EDI principles. Instead of a standard slideshow, I lead the technical implementation of a custom RAG (Retrieval-Augmented Generation) Chatbot.
The bot acts as a virtual coach, answering student questions based on a curated knowledge base of academic resources, class notes, and research papers.
- Engine: Botpress (NLP & Knowledge Base)
- Frontend: React (Vite/Next.js)
- Styling: Tailwind CSS (Custom University Palette)
- Hosting: Vercel
The bot doesn't just hallucinate answers; it follows a strict data pipeline:
- Curated Data: We scraped authoritative websites on teamwork dynamics and digitized a semester's worth of class notes.
- Knowledge Base: Data was cleaned, condensed, and uploaded to Botpress vector stores.
- Prompt Engineering: The system prompt enforces a specific 4-step response algorithm:
- Identify: Acknowledge the specific student concern.
- Connect: Link the issue to a pillar of EDI.
- Propose: Offer concrete, actionable solutions.
- Follow Up: Ask a question to keep the user engaged.
The Session Persistence Bug Integrating the Botpress webchat into a custom React UI caused session issues. The bot would create a new session on every page reload, wiping context.
Solution: I implemented custom session management logic in the React useEffect hooks to check for existing conversation IDs in localStorage before initializing a new connection. This ensured conversation history persisted across reloads.