ALAITAPTAP

Walking navigation for Metro Manila that prefers safer streets, not just the fastest path. The app combines crime incident data, lighting and time-of-day context, safe spots (e.g. lit areas, police, commercial corridors), and an ML-based safety scorer so users can plan routes with clearer situational awareness—especially women walking alone.

Built for the AI ASEAN Youth Challenge 2026.

What we’re building

• Map + route planner — Origin/destination, map visualization (Leaflet), and routes biased by safety scores along the graph.
• Backend API — FastAPI service for routing (OpenRouteService), per-segment safety scoring, crime/safe-spot data, and optional SMS (Twilio) for sharing location or alerts.
• Data layer — PostgreSQL in production (SQLite supported for local dev); seeded crime incidents and safe spots for realistic demos.

Repository layout

Path	Role
saferoute-backend/	FastAPI app (/api routing, safety, spots), SQLAlchemy models, ML scorer, Twilio helper, seed scripts
saferoute-frontend/	React 19 + Vite + React Router + Leaflet UI

Tech stack

Layer	Choices
API	Python 3, FastAPI, Uvicorn
Data	SQLAlchemy, PostgreSQL (prod) / SQLite (dev)
Routing	OpenRouteService
ML	scikit-learn–based safety scoring
Notifications	Twilio SMS (optional)
Web	React, Vite, Leaflet / react-leaflet, Axios

Local development

Step-by-step (including empty VITE_API_URL for the dev proxy) is in LOCAL_DEVELOPMENT.md. That file also explains how to revert to production (Vercel VITE_API_URL, redeploy) when local work is done.

Backend — from saferoute-backend/:

1. Create a virtual environment and install dependencies: pip install -r requirements.txt
2. Copy saferoute-backend/.env.example to saferoute-backend/.env and set DATABASE_URL, OPENROUTESERVICE_API_KEY, and Twilio vars if you use SMS.
3. Seed the database (CSV files are in saferoute-backend/data/, or regenerate with python data/scrape_crime_data.py and python data/fetch_safe_spots.py):
   python scripts/seed_database.py
4. Start the API: uvicorn app.main:app --reload (from saferoute-backend/).

Frontend — from saferoute-frontend/:

1. npm install
2. Copy .env.example to .env. For local dev, leave VITE_API_URL empty so Vite proxies /api to localhost:8000, or set VITE_API_URL to your API (Railway URL for production builds on Vercel).
3. npm run dev (default Vite dev server, often http://localhost:5173).

Deployment (target)

• API: Railway — set Root Directory to saferoute-backend, add the PostgreSQL plugin (injects DATABASE_URL), set env vars from .env.example, Start Command is defined in saferoute-backend/railway.toml (uvicorn app.main:app --host 0.0.0.0 --port $PORT). After first deploy, run once: python scripts/seed_database.py (Railway CLI shell or one-off job) so Postgres has schema + seed data.
• Web: Vercel — set Root Directory to saferoute-frontend, add VITE_API_URL = your Railway API URL — production: saferoute-asean.vercel.app.

Tune CORS_ORIGINS in the backend env to include https://saferoute-asean.vercel.app (and localhost for dev).

Phase 1 deliverables (checklist)

Item	How to verify
React on Vercel	Load saferoute-asean.vercel.app; map loads.
PostgreSQL + schema	Railway Postgres attached; tables created on app boot (Base.metadata.create_all).
500+ crime incidents	data/crime_incidents.csv has 520 rows; seed_database.py loads them.
50+ safe spots	data/safe_spots.csv has 55 rows; seeded with script.
OpenRouteService	OPENROUTESERVICE_API_KEY set in Railway (and local .env).
Twilio SMS	python scripts/verify_twilio.py (optional: TWILIO_SEND_TEST=1 + TWILIO_TEST_TO).
ML safety scorer	GET /api/safety-score?lat=14.65&lng=121.07&hour=22 returns score + label; night vs day differs.
GitHub clean commits	Feature commits on main; push to origin.

License / data

Crime and POI data are for demonstration; production use requires appropriate sources, licenses, and privacy review.