F26 · COMP 523

Library Navigator

An AI-assisted library chatbot helping rural North Carolina communities find and use the resources their libraries already pay for.

What is it?

Library Navigator is an AI-assisted chatbot that helps people in rural North Carolina find and actually use the library resources their county already pays for.

A patron types a ZIP code and a question in plain English, such as how to find reliable information about managing diabetes. The assistant answers with the specific resources their own library system offers, ranked by how close they are, and then walks them through getting in: which site to open, where the login is, what their library card number does, and what to click next.

Why it matters

Public libraries across the state subscribe to far more than most people ever use. Health databases, e-book and audiobook services, tutoring, job tools, and local programs are all sitting there unused. The problem is not access. The problem is that almost nobody knows any of it exists.

This matters most in rural counties, where the library may be the only reliable on-ramp to health information and study support, and where patrons are least likely to have someone to ask. It is not only patrons who are in the dark. Librarians themselves often cannot name the full catalog of subscriptions their system carries.

How it works

One
Accurately sourced and cited data
Answers come from a curated dataset of roughly 467 resource records covering all 100 North Carolina counties, 90 library systems, and 413 outlets. The assistant is not improvising.
Two
Interactive guided assistants
Modeled on the NC DMV AI agent, the assistant walks people through the real access steps one at a time instead of dropping a link and leaving.
Three
Honest about uncertainty
High-confidence answers are served directly. Medium and low confidence answers are flagged, cited, and routed to a human librarian. The assistant is allowed to say it does not know.
Four
Built to be maintained
Librarians can add, edit, override, and retire entries, so the data does not go stale the day the semester ends.

This semester

The deliverable is a functioning proof of concept with working code, not a clickable mockup. The first iteration uses the curated dataset as its backend. The longer-term vision, beyond this semester, is a live agent that keeps the database current on its own under librarian oversight.

Read the full specification, or see who is building it on the team page.