(OUT OF DATE) Parkaway (Smart Parking Recommender for Kentron District in Yerevan, Armenia)

A Flask + Leaflet.js web app that turns your original Folium/OSMnx script into a live, interactive parking dashboard for Kentron, Yerevan.

What's inside

yerevan-smart-parking/
├── app.py                  # Flask backend: data loading, simulation, API
├── requirements.txt
├── templates/
│   └── index.html          # Leaflet dashboard (sidebar + map, single file)
└── data/
    └── parking_cache.geojson   # created automatically on first run

How it works

• Data source (app.py): on startup, the app tries to pull live curb/lot parking geometry for "Kentron, Yerevan, Armenia" via osmnx.features_from_place, using the same tag set and red/green/blue/grey classification logic as your original script. A successful pull is cached to data/parking_cache.geojson so future restarts are instant. If OSMnx/Overpass isn't reachable (no network, rate-limited, etc.), the app automatically falls back to a curated offline dataset covering well-known Kentron corridors (Northern Avenue, Mashtots Avenue, Republic Square, the Opera lot, and more), so the app is always demoable.
• Simulated real-time layer: simulate_segment_state() models congestion as a function of hour-of-day and weekday/weekend, with morning (08:00–10:00) and evening (17:00–20:00) rush-hour peaks in Kentron and low congestion at night, plus a small deterministic per-street "personality" offset. This drives availability_pct, occupancy_status, and traffic_level for every segment, refreshed every minute.
• Recommendation engine (/api/recommend): given a destination lat/lon, it finds every parking segment within radius (default 500 m), scores it as type_weight - traffic_penalty - distance_penalty (free > lot > paid > unspecified, penalized by live congestion and distance), and returns the top 3.

Install & run

cd yerevan-smart-parking
python3 -m venv venv && source venv/bin/activate   # optional but recommended
pip install -r requirements.txt
python app.py

Then open http://127.0.0.1:5000 in your browser.

The first launch will try to fetch live OSM data (can take 10–30s); if that fails for any reason, you'll see a console log and the app seamlessly starts with the offline dataset instead — no action needed.

Using the dashboard

• Click anywhere on the map (or pick a landmark preset — Republic Square, Opera House, Cascade, Vernissage) to drop a destination pin.
• The "Top 3 Recommended Spots" panel updates instantly, each card showing a live availability gauge, walking distance/time, and score.
• The "Live Traffic" panel lists every street ranked by current simulated congestion, refreshing every 15 seconds along with the map layer.
• Click any street or lot on the map for a popup with its live status.

API reference

Endpoint	Method	Params	Returns
/api/parking	GET	–	GeoJSON FeatureCollection of all segments with live parking_type, occupancy_status, availability_pct, congestion_score, traffic_level
/api/recommend	GET	lat, lon, radius (m, default 500)	Top 3 scored parking recommendations near a destination
/api/traffic	GET	–	Street-by-street congestion snapshot + overall Kentron traffic level
/api/refresh	POST	–	Forces a fresh OSMnx pull (falls back to cache/offline data on failure)

Notes for production use

• Swap the simulated availability model for a real feed (municipal sensors, a partner API, or crowdsourced check-ins) by replacing the body of simulate_segment_state() — the rest of the app (API contract, frontend) doesn't need to change.
• For heavier traffic, put PARKING_FEATURES behind a proper cache (Redis) instead of the in-memory list, and move the OSMnx refresh to a scheduled background job rather than a request-triggered one.