Nahvi

Nahvi is a static, privacy-preserving agentic mapping assistant for accessibility, environmental health, disaster resilience, and equitable urban planning.

• Hosting target: GitHub Pages
• Rendering: Leaflet (+ Leaflet.heat)
• Spatial operations: Turf.js
• Routing / isochrones / optimization: OpenRouteService (ORS)
• Data model: Local static GeoJSON overlays + optional live OpenAQ fetch

Mission

Nahvi helps answer questions such as:

• What is the safest and most accessible route to a destination?
• Which corridors have lower pollution or heat stress?
• Which neighborhoods are cut off by hazard zones?
• What amenities are reachable within 15 minutes?

Project structure

nahvi/
├── index.html
├── css/
│   └── style.css
├── js/
│   ├── map.js
│   ├── routes.js
│   ├── isochrones.js
│   ├── heatmaps.js
│   ├── dataLoader.js
│   └── ui.js
├── data/
│   ├── amenities.geojson
│   ├── hazard_zones.geojson
│   └── sample_crime.geojson
└── .github/workflows/
    └── update_data.yml

Agentic tool interface implemented

The app includes frontend “tool” functions that the client-side assistant can call:

• calculate_isochrone(location, time, profile) in /js/isochrones.js
• find_safe_route(start, end, profile) in /js/routes.js
• query_overlay_data(bounding_box, layer_type) in /js/dataLoader.js
• render_geojson_on_map(geojson, style_options) in /js/map.js

BYO-key security model

Nahvi stores user-provided API keys in browser localStorage only:

• nahvi_ors_key
• nahvi_openai_key (optional)
• nahvi_anthropic_key (optional)

No backend persistence or tracking is implemented.

Features included now

• Wheelchair-capable ORS routing profile toggle
• Hazard-zone route avoidance option using ORS avoid_polygons
• 5–60 minute ORS isochrones with amenity counting (15-minute city baseline)
• Safety heatmap from static crime sample
• Air quality overlay from OpenAQ API
• Data-source link display for each local overlay
• “Use my location” geolocation shortcut
• Volunteer stop optimization demo (ORS optimization API)
• Lightweight prompt box that maps intent keywords to tool actions

Running locally

Because the app uses ES modules and fetches local GeoJSON files, serve it over HTTP:

cd path/to/nahvi
python3 -m http.server 8080

Then open http://localhost:8080.

Data transparency and ethics

• Overlay sources are public/open (OSM, FEMA-style hazard examples, city open data style samples).
• Nahvi is a planning tool and not a replacement for official emergency alerts.
• ORS wheelchair routing is useful but should be verified against ground truth.
• ORS free-tier limits apply; user-facing errors are surfaced when requests fail.