Bot-BU

Inspiration

Info lives in many places, BU Brain, emails, PDFs, and flyers. Students just want one place to ask and get a clear answer. We built a chat that feels like asking the campus itself.

What it does

Bot-BU is an intelligent chat for Binghamton University. It gives exclusive university updates about:

• Academics: policies, deadlines, advising, classes
• Campus events: what’s happening today and this week
• Real-time info: dining hours, buses, alerts, and more

Answers come from official BU sources and live feeds where possible.

How we built it

• Frontend: Next.js
• UI: Tailwind CSS
• Search: chunking + embeddings in a vector store for fast, relevant lookup
• AI: Google Generative AI (Gemini) for clear, grounded replies
• Hosting: Vercel

Challenges we ran into

• Started with linear keyword search. It was slow and missed context.
• Switched to embeddings + vector search. Much faster and more accurate.
• PDF text is messy. Needed cleanup for tables, headers, and footers.
• Managed caching, rate limits, and kept answers tied to sources.

Accomplishments we’re proud of

• A simple chat that understands campus questions.
• Source-backed answers with short, clear replies.
• Works well on phone and laptop.
• Mixes real-time data with verified policy docs.

What we learned

• Vector search beats linear search for campus content.
• Chunk size and prompting matter a lot.
• Users want quick answers first, links for details.
• Streaming and caching make the app feel fast.

What’s next for Bot-BU

• Kiosk devices across campus (Union, libraries, dorm lobbies) so students can walk up and ask.
• Deeper live data: events, dining, buses, safety alerts.
• Multi-document chat memory and summaries.
• Voice mode and multilingual support.
• Admin tools to curate sources and track gaps.

Built With

• nextjs
• pinecone
• tailwindcss