SafeWalk NYC 🗽

AI-powered pedestrian safety routing for New York City — built for students walking at night.

SafeWalk NYC combines real-time crime data, streetlight complaints, traffic signal density, and live news alerts to score walking routes 0–100. Claude AI explains the recommendation in plain English so users make informed decisions, not just fast ones.

Demo

User: "Walk me from 3009 Broadway to Times Square"

┌─────────────────────────────────────────────────────────────────┐
│  3 ROUTES FOUND              RECOMMENDED: Route 1               │
│                                                                 │
│  Route 1  ★ Best    98/100 Safe    via Broadway      3.97 mi    │
│  Route 2            98/100 Safe    via Central Park  4.74 mi    │
│  Route 3            97/100 Safe    via Riverside     4.43 mi    │
│                                                                 │
│  "Route 1 via Broadway is your best option — zero crime         │
│   incidents, shortest at 3.97 mi, well-traveled corridor."     │
└─────────────────────────────────────────────────────────────────┘

All 3 routes appear on the Mapbox map simultaneously, each color-coded by safety score with floating score badges.

Features

Conversational input — describe your destination naturally; Claude parses intent into street addresses
3 route alternatives — direct path + two corridor variants (east/west) scored independently
Live safety data — NYPD crime reports (90-day window), NYC 311 lighting complaints, traffic signal density
Real-time news — Linkup API surfaces construction closures and safety alerts near your route
Score badges on map — all 3 routes visible simultaneously with floating score chips
Per-route cards — each card shows its distinguishing corridor, Crime/Lighting/Signal bars, and an expandable explanation
Time-aware scoring — same street scores lower at 2 AM (×2.0) than at noon (×1.0)
Campus safety alerts — "I feel unsafe" button routes to Columbia, NYU, Barnard, or 911

Architecture

Full Request Pipeline

User message  (e.g. "Walk me from 3009 Broadway to Times Square")
     │
     ▼
┌──────────────────────────────────────┐
│  Step 1 — Claude: Parse Intent       │
│                                      │
│  Extracts origin + destination as    │
│  precise street addresses.           │
│  "Columbia Biz School" → "665 W 130th St"
└───────────────────┬──────────────────┘
                    │
                    ▼
┌──────────────────────────────────────┐
│  Step 2 — Geocoding (parallel)       │
│                                      │
│  Mapbox Geocoding v5                 │
│  • types: address, poi, neighborhood │
│  • bbox: NYC boroughs only           │
│  • NYC coordinate validation         │
│                                      │
│  Nominatim (OSM) fallback            │
│  • handles institutional POIs        │
│  • viewbox: NYC, bounded=1           │
└───────────────────┬──────────────────┘
                    │  origin + dest coordinates
                    ▼
┌──────────────────────────────────────────────────────────┐
│  Step 3 — Mapbox Directions  (3 parallel requests)       │
│                                                          │
│  Request A:  direct route          (no waypoint)         │
│  Request B:  eastern variant       (mid + perp × 0.012°) │
│  Request C:  western variant       (mid − perp × 0.012°) │
│                                                          │
│  Deduplicate by 2nd-most-prominent street + distance     │
│  → up to 3 distinct MapboxRoute objects                  │
└──────────────────────────┬───────────────────────────────┘
                           │
                           ▼
┌──────────────────────────────────────────────────────────┐
│  Step 4 — Segment Splitting  (~200 m chunks)             │
│                                                          │
│  Each Mapbox step → RouteSegment { start, end, geom }    │
│  Shared endpoint deduplication                           │
└──────────────────────────┬───────────────────────────────┘
                           │
                           ▼
┌──────────────────────────────────────────────────────────┐
│  Step 5 — Data Enrichment  (per segment, all parallel)   │
│                                                          │
│   ┌─────────────────┐  ┌─────────────────┐  ┌─────────┐ │
│   │  NYPD Crimes    │  │  311 Lighting   │  │ Traffic │ │
│   │  250 m radius   │  │  150 m radius   │  │ Signals │ │
│   │  90 days back   │  │  open only      │  │ 150 m   │ │
│   │  5uac-w243      │  │  erm2-nwe9      │  │ e5mv-wy8│ │
│   └─────────────────┘  └─────────────────┘  └─────────┘ │
│                                                          │
│   + Linkup news (once per route, at route midpoint)      │
│     → construction closures, crime alerts, detours       │
└──────────────────────────┬───────────────────────────────┘
                           │
                           ▼
┌──────────────────────────────────────────────────────────┐
│  Step 6 — Safety Scoring                                 │
│                                                          │
│  Per segment:                                            │
│    crimePenalty    = Σ(crime × severity) [FELONY=10]    │
│    lightPenalty    = openComplaints × 8  [cap 40]       │
│    signalPenalty   = 0/5/15 for ≥2/1/0 signals          │
│    newsPenalty     = Σ severity         [cap 15]        │
│                                                          │
│    raw    = crime×0.35 + light×0.25                      │
│             + signal×0.15 + news×0.10                    │
│    score  = 100 − raw × timeModifier                     │
│                                                          │
│  timeModifier:  06–20h ×1.0 · 20–24h ×1.5               │
│                 00–04h ×2.0 · 04–06h ×1.3               │
│                                                          │
│  Route score = distance-weighted avg of segments         │
└──────────────────────────┬───────────────────────────────┘
                           │
                           ▼
┌──────────────────────────────────────────────────────────┐
│  Step 7 — Claude: Explanation                            │
│                                                          │
│  Summarises all 3 routes, recommends the best by name,  │
│  cites crime counts, lighting issues, worst block.       │
└──────────────────────────┬───────────────────────────────┘
                           │
                           ▼
                    API Response
          { routes[], recommended, explanation }

Safety Score Breakdown

                  Crime (35%)        Lighting (25%)
                      │                   │
        NYPD data ────┤        311 data ──┤
        90 days       │        open only  │
        250 m         │        150 m      │
                      ▼                   ▼
                 crimePenalty        lightPenalty
                      │                   │
                      └────────┬──────────┘
                               │
              Signal (15%) ────┤◄── Traffic signals 150 m
              News   (10%) ────┤◄── Linkup alerts/closures
                               │
                               ▼
                    raw = weighted sum
                               │
                               ▼
                   score = 100 − raw × T
                               │
               ┌───────────────┼───────────────┐
               ▼               ▼               ▼
           T = 1.0         T = 1.5         T = 2.0
          (6 AM–8 PM)    (8 PM–midnight)  (midnight–4 AM)

Score labels:  ≥ 80  🟢 Safe
               ≥ 60  🟡 Moderate
               ≥ 40  🟠 Use Caution
                < 40  🔴 High Risk

Frontend Data Flow

route-planner/page.tsx
        │
        │  User types destination
        ▼
ChatInput.tsx ──POST /api/chat──► Claude (conversational)
        │                               │
        │  onRouteExtracted(org, dst)   │ plain-text reply
        ▼                               ▼
handleRouteExtracted()          Chat bubble in sidebar
        │
        │  POST /api/score-route
        ▼
ApiResponse { routes[], recommended, explanation }
        │
        ├──► MapView.tsx  (Mapbox GL JS, SSR: false)
        │     • All 3 polylines drawn simultaneously
        │     • Color by score: green / yellow / orange / red
        │     • Selected route: opacity 1.0, width 6 px
        │     • Others: opacity 0.45, width 3 px
        │     • Score badge (pill) at each route midpoint
        │     • Origin (blue dot) + Destination (red dot)
        │     • Camera fits all routes on load
        │
        └──► RouteCard.tsx  ×3  (one per route)
              • "Route N  ·  via [street]"
              • ★ Best badge on recommended
              • Safety score + Shield icon
              • Duration · Distance · Incidents
              • Crime / Lighting / Signals progress bars
              • Expandable explanation
                - Recommended → full Claude narrative
                - Others      → auto-generated from data

File Structure

safewalk-nyc/
├── app/
│   ├── page.tsx                    Landing page with email signup
│   ├── layout.tsx                  Root layout — loads globals.css
│   ├── globals.css                 Mapbox CSS + Tailwind directives
│   ├── route-planner/
│   │   └── page.tsx                Main app: map + chat sidebar
│   └── api/
│       ├── score-route/route.ts    POST — full 7-step scoring pipeline
│       ├── chat/route.ts           POST — conversational assistant
│       └── alert/route.ts          POST — campus safety alert routing
│
├── components/
│   ├── MapView.tsx                 Mapbox GL map (dynamic import, SSR false)
│   ├── RouteCard.tsx               Per-route card with safety bars
│   ├── ChatInput.tsx               Conversational input + message history
│   ├── AlertButton.tsx             Persistent "I feel unsafe" button
│   └── SignupForm.tsx              Email signup with university detection
│
├── lib/
│   ├── types.ts                    All shared TypeScript interfaces
│   ├── scoring.ts                  scoreSegment() · scoreRoute() · scoreLabel()
│   ├── alert-routing.ts            Email domain → campus safety contacts
│   └── api/
│       ├── nyc-open-data.ts        NYPD · 311 · traffic signal fetchers
│       ├── linkup.ts               Linkup search + news penalty calculator
│       └── claude.ts               Anthropic SDK wrappers
│
├── .env.local.example              Required environment variables template
├── CLAUDE.md                       Project conventions (AI-assisted coding)
└── README.md

Tech Stack

Layer	Technology
Framework	Next.js 14 — App Router, TypeScript
Styling	Tailwind CSS
Map	Mapbox GL JS (dynamic import, SSR disabled)
AI	Anthropic Claude `claude-sonnet-4-6`
Crime data	NYC Open Data — NYPD Complaints `5uac-w243`
Lighting data	NYC Open Data — 311 Service Requests `erm2-nwe9`
Signal data	NYC Open Data — Traffic Signals `e5mv-wy8f`
News alerts	Linkup API
Geocoding	Mapbox Geocoding v5 + Nominatim (OSM) fallback
Routing	Mapbox Directions v5 — 3 variants via forced waypoints

Getting Started

1. Clone and install

git clone https://github.com/Tanmay244/safewalk-nyc.git
cd safewalk-nyc
npm install

2. Configure environment variables

cp .env.local.example .env.local

Edit .env.local:

ANTHROPIC_API_KEY=sk-ant-...          # console.anthropic.com
LINKUP_API_KEY=...                    # linkup.so
NEXT_PUBLIC_MAPBOX_TOKEN=pk.eyJ...    # mapbox.com

3. Run dev server

npm run dev
# → http://localhost:3000

Navigate to /route-planner and try:

Walk me from 3009 Broadway to Times Square

You should see 3 routes on the map with safety score badges.

API Reference

`POST /api/score-route`

Score 3 walking route alternatives between two NYC addresses.

Request body

{
  "message": "Walk me from 3009 Broadway to Times Square",
  "userLocation": { "lat": 40.8095, "lng": -73.9634 }
}

Response

{
  "routes": [
    {
      "path": { "type": "LineString", "coordinates": [[...]] },
      "score": 98,
      "scoreLabel": { "label": "Safe", "color": "green" },
      "distanceMiles": 3.97,
      "durationMinutes": 74,
      "label": "via Broadway",
      "segments": [
        {
          "streetName": "Broadway",
          "score": 98,
          "crimePenalty": 0,
          "lightingPenalty": 4,
          "trafficSignalPenalty": 0,
          "crimeCount": 0,
          "openLightingIssues": 1,
          "nearbySignals": 3
        }
      ]
    }
  ],
  "recommended": 0,
  "explanation": "Route 1 via Broadway is your best option...",
  "origin": { "lat": 40.8095, "lng": -73.9634 },
  "destination": { "lat": 40.7580, "lng": -73.9855 },
  "timeOfDay": 13
}

`POST /api/chat`

Conversational assistant for safety questions and route queries.

Request

{
  "messages": [{ "role": "user", "content": "Is it safe to walk at midnight?" }],
  "context": { "currentLocation": "Columbia University" }
}

Response

{
  "reply": "Walking late at night has some risks worth knowing...",
  "route": { "origin": "Columbia University", "destination": "Penn Station" }
}

`POST /api/alert`

Route a safety alert to campus security or emergency contacts.

Request

{
  "email": "student@columbia.edu",
  "location": { "lat": 40.8075, "lng": -73.9626 },
  "message": "I feel unsafe near 120th and Amsterdam"
}

University Safety Contacts

Email domain detection automatically routes alerts to the right team.

University	Domain	Emergency
Columbia University	`columbia.edu`	212-854-5555
NYU	`nyu.edu`	212-998-2222
Barnard College	`barnard.edu`	Barnard Security
The New School	`newschool.edu`	Campus Safety
Fordham University	`fordham.edu`	Fordham Security

Adding a university — append to UNIVERSITIES in lib/alert-routing.ts:

{ name: "My University", domain: "myuni.edu", safetyEmail: "safety@myuni.edu", color: "#3b82f6" }

Score Reference

Score	Label	Meaning
80–100	🟢 Safe	Well-lit, low crime, good pedestrian signals
60–79	🟡 Moderate	Some concerns; take normal precautions
40–59	🟠 Use Caution	Elevated risk; stay aware of surroundings
0–39	🔴 High Risk	Avoid if possible, especially at night

Scores are time-adjusted — the same block is scored lower after dark:

06:00 – 20:00  ×1.0  (baseline)
20:00 – 24:00  ×1.5  (evening penalty)
00:00 – 04:00  ×2.0  (night penalty)
04:00 – 06:00  ×1.3  (pre-dawn penalty)

Roadmap

Real-time MTA transit alerts → route avoidance
Email delivery for safety alerts (Resend / SendGrid)
Persist signups to Supabase
Rate limiting via @upstash/ratelimit
User accounts + saved routes
Mobile app (React Native / Expo)

Contributing

Pull requests welcome. For major changes, open an issue first.

License

MIT — built for NYC students who deserve to feel safe getting home.

Name		Name	Last commit message	Last commit date
Latest commit History 5 Commits
app		app
components		components
lib		lib
.env.local.example		.env.local.example
.eslintrc.json		.eslintrc.json
.gitignore		.gitignore
CLAUDE.md		CLAUDE.md
LICENSE		LICENSE
README.md		README.md
next.config.mjs		next.config.mjs
package-lock.json		package-lock.json
package.json		package.json
postcss.config.mjs		postcss.config.mjs
tailwind.config.ts		tailwind.config.ts
tsconfig.json		tsconfig.json

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

SafeWalk NYC 🗽

Demo

Features

Architecture

Full Request Pipeline

Safety Score Breakdown

Frontend Data Flow

File Structure

Tech Stack

Getting Started

1. Clone and install

2. Configure environment variables

3. Run dev server

API Reference

`POST /api/score-route`

`POST /api/chat`

`POST /api/alert`

University Safety Contacts

Score Reference

Roadmap

Contributing

License

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Folders and files

Latest commit

History

Repository files navigation

SafeWalk NYC 🗽

Demo

Features

Architecture

Full Request Pipeline

Safety Score Breakdown

Frontend Data Flow

File Structure

Tech Stack

Getting Started

1. Clone and install

2. Configure environment variables

3. Run dev server

API Reference

POST /api/score-route

POST /api/chat

POST /api/alert

University Safety Contacts

Score Reference

Roadmap

Contributing

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

`POST /api/score-route`

`POST /api/chat`

`POST /api/alert`

Packages