RPInSight

AI-assisted campus navigation with Mapbox, GeoJSON, and curated RPI location data.

RPInSight is a campus navigation prototype for Rensselaer Polytechnic Institute. It combines a Mapbox campus map, curated GeoJSON location layers, directions, and a campus-grounded search assistant for finding dining, study, lecture, and parking locations.

Quick nav: Demo Scope · Features · Data · Architecture · Security · Quick Start · Validation

Demo Scope

RPInSight is supporting evidence for my QGIS/geospatial portfolio. It is not the main QGIS proof; the flagship asset package lives here:

https://github.com/Taz33m/qgis-ai-geospatial-assets

This repo shows the adjacent product layer: turning curated location data into a usable map interface with route UX, search, and location-aware responses.

Core Features

• Mapbox campus map: interactive RPI campus navigation with custom style support.
• Curated GeoJSON layers: dining, study, lecture, and parking points stored under src/data/.
• Campus search assistant: /api/search grounds answers in the local campus dataset and returns a building name, description, coordinates, and confidence cue.
• Directions API route: /api/directions proxies Mapbox Directions requests without logging token-bearing URLs.
• Route rendering: walking, cycling, and driving paths are drawn back onto the map.
• Data QA panel: the UI links this product back to the public QGIS QA portfolio.

Product Flow

flowchart LR
  User["Campus query"] --> Search["/api/search"]
  Search --> Data["Curated GeoJSON layers"]
  Search --> Answer["Building + description + confidence"]
  Answer --> Map["Mapbox marker / fly-to"]
  User --> Location["Optional geolocation"]
  Location --> Directions["/api/directions"]
  Directions --> Route["Rendered route layer"]

Loading

Campus Data

Layer	File	Count	Purpose
Dining halls	src/data/dining_halls.geojson	4	Meal-plan dining and food locations.
Study halls	src/data/study_halls.geojson	4	Library and student study spaces.
Lecture halls	src/data/lecture_halls.geojson	9	Major classroom and academic buildings.
Parking	src/data/parking.geojson	5	Visitor and permit parking points.

The campus dataset is intentionally compact and reviewable. Search confidence is explicit: exact matches use curated coordinates, while uncertain results are marked approximate or withheld.

Architecture

RPInSight is a Next.js application with three small layers:

• UI layer: src/components/MapComponent.jsx, SearchBar.tsx, and supporting UI components.
• Data layer: bundled GeoJSON feature collections in src/data/.
• API layer: server routes for campus search and Mapbox directions under src/app/api/.

src/app/
  api/directions/route.ts
  api/search/route.ts
src/components/
  MapComponent.jsx
  SearchBar.tsx
  DataMethodologyPanel.tsx
src/data/
  dining_halls.geojson
  lecture_halls.geojson
  parking.geojson
  study_halls.geojson

Security and Privacy

This repo has been cleaned up for public presentation:

• No committed API keys or secret .env files.
• Public Mapbox configuration is read from NEXT_PUBLIC_* variables; server routes can use server-only MAPBOX_TOKEN.
• API routes validate request bodies before calling external services.
• Search and directions routes include lightweight request throttling.
• Directions are bounded to the RPI campus area to reduce accidental token abuse.
• Directions requests do not log full Mapbox URLs or access tokens.
• Popup HTML escapes GeoJSON properties before rendering.
• Security headers disable framing, MIME sniffing, and unnecessary browser permissions.

See SECURITY.md for the public security posture and reporting note.

Quick Start

cp .env.example .env.local
npm install
npm run dev

Open http://localhost:3000.

Required environment variables:

• NEXT_PUBLIC_MAPBOX_TOKEN: public Mapbox token for browser map rendering.
• NEXT_PUBLIC_MAPBOX_STYLE_URL: Mapbox style URL.
• NEXT_PUBLIC_CAMPUS_CENTER_LNG / NEXT_PUBLIC_CAMPUS_CENTER_LAT: default map center.
• MAPBOX_TOKEN: optional server-side Mapbox token for directions.
• OPENAI_API_KEY: optional server-side key for AI-backed campus search. Without it, exact local dataset matches still work.

Validation

npm run type-check
npm run lint
npm run build
npm audit --audit-level=moderate

Limitations

• RPInSight is not an official RPI service.
• The dataset is curated for a product demo, not exhaustive campus operations.
• Directions depend on Mapbox coverage and the provided style/token configuration.
• AI search should be treated as campus assistance, not authoritative emergency or accessibility guidance.

Portfolio Context

Recommended submission hierarchy:

1. Main portfolio page: https://tazeemm.com/qgis-portfolio
2. Flagship QGIS asset package: https://github.com/Taz33m/qgis-ai-geospatial-assets
3. Companion evaluator toolkit: https://github.com/Taz33m/gis-ai-evaluation-lab
4. Supporting geospatial product: this repo.