Demo:

https://youtu.be/RcQGiWM8hWc

Weather Forecast

A full-stack weather forecasting application with Google OAuth, statistical analysis, anomaly detection, and location-based predictions across 600 cities.

Also available in Romanian — see documentation/referat_prognoza_meteo.docx and arhitectura.md.

---

Stack

Layer	Technology
Frontend	Next.js, React 19, TypeScript, Vite, Recharts, Leaflet, React-Leaflet, Axios
Backend	Rust, Axum 0.8, SQLx, Tokio, jsonwebtoken, reqwest, tower-http
Database	PostgreSQL with 18 SQL migrations (PL/pgSQL functions, triggers, stored procedures)
Scripting	Python (seed, alerts, anomaly generation)

---

Data

• ~5 million rows inserted across the database
• 600 cities from around the world (40 per country, balanced by population)
• 853 days of historical weather data (April 2024 – May 2026) sourced from Open-Meteo API
• 18 database migrations implementing business logic directly in PostgreSQL

---

Architecture

PROJECT/
├── backend/          # Rust + Axum REST API
│   ├── src/
│   │   ├── main.rs           # Server startup, CORS, migrations
│   │   ├── db.rs             # PgPool connection (max 10 connections)
│   │   ├── routes.rs         # 35+ public & protected endpoints
│   │   ├── handlers/         # Business logic layer
│   │   │   ├── auth.rs       # Google OAuth, JWT, refresh tokens
│   │   │   ├── forecasts.rs  # Predictions, bulk predictions, history
│   │   │   ├── discovery.rs  # City/country search, map bounds query
│   │   │   ├── social.rs     # Comments, reactions (like/dislike)
│   │   │   ├── stats.rs      # Country dashboards, clusters, leaderboards
│   │   │   ├── user.rs       # Profile, favorites, settings
│   │   │   └── weather.rs    # Current weather, alerts, history
│   │   └── models/           # DTOs & DB mappings
│   └── migrations/           # 18 SQL migrations (see below)
├── frontend/         # Next.js + React 19
│   └── src/
│       ├── pages/            # Dashboard, CityDetails, Statistics, Favorites, Settings, Login
│       ├── components/       # MapView, WeatherCard, Navbar, AlertBanner, ProtectedRoute
│       ├── services/         # Axios with auto token refresh & request queuing
│       ├── context/          # AuthContext (user state, login/logout)
│       ├── hooks/            # useAuth
│       └── types/            # TypeScript interfaces
├── scripts/          # Python utilities
│   ├── seed_meteo.py         # Seeds 5M rows from Open-Meteo API
│   ├── generate_alerts.py    # Injects extreme weather data to trigger DB alerts
│   └── generate_romania_anomaly.py  # Generates temperature anomalies for testing
└── worldcities.csv           # Data source (population-balanced city selection)

---

Cool Features

1. Prediction Engine (PL/pgSQL)

get_city_prediction() at backend/migrations/002_create_prediction_function.sql combines:

• 7-day recent trend offset capped at ±4°C
• DOY window averages (±3 days around target date across all years)
• 3-year historical extrapolation — compares same day-of-year in prior years against their window averages to compute deltas
• Result is clamped to realistic ranges

get_city_prediction_bulk() wraps this to return up to 10 days at once, pre-decorated with icons and UV index.

2. Bulk Prediction with Concurrent Fetching

The frontend (CityDetails.tsx) fetches predictions in parallel batches of 6 for missing dates. For 5-day, 7-day, 10-day ranges it uses the bulk endpoint; for 1-month and 1-year ranges it selectively fetches missing dates only.

3. Weather Icon Classification

generate_weather_icon() at backend/migrations/008_generate_weather_icon.sql classifies conditions into 12 types: ☀️ Senin → ☀️ Caniculă → 🔥 Caniculă cu umiditate → ⛅ Parțial înnorat → ☁️ Înnorat → 🌦️ Ploaie ușoară → 🌧️ Ploaie abundentă → 🌬️ Vânt puternic cu ploaie → 💨 Vânt puternic → ⛈️ Furtună → ❄️ Ninsoare → 🥶 Ger

Automatically populated via a BEFORE INSERT OR UPDATE trigger on the forecasts table.

4. UV Index Generation

generate_uv_index() at backend/migrations/012_add_uv_index_generation.sql derives UV levels (Low → Moderate → High → Very High → Extreme) from temperature, humidity, and wind speed. Also auto-filled via trigger on every forecast insert/update.

5. Extreme Weather Alert Trigger

check_extreme_weather() at backend/migrations/003_create_weather_alerts.sql is an AFTER INSERT trigger that fires on every new forecast row and creates alerts if:

• temp_max ≥ 35°C → extreme heat alert with recommendations
• temp_min ≤ -10°C → extreme cold alert
• wind_speed ≥ 50 km/h → high wind alert
• humidity ≥ 95% → high humidity alert

Duplicate alerts are prevented (checks if forecast_id already has an alert).

6. Comment System with Anti-Spam & Reputation

• Votes: Like/dislike on comments, supports both authenticated users and anonymous guest tokens (x-guest-token header)
• Anti-spam: trg_handle_vote_antispam() at backend/migrations/015_add_vote_antispam_and_icon_trigger.sql enforces a 3-second cooldown between votes using a reaction_logs table
• Reputation system: adjust_comment_vote_reputation() at backend/migrations/011_add_comment_vote_reputation.sql adjusts comment author's reputation on each vote:
  • Users with reputation ≥ 100 have double voting power (20 instead of 10)
  • Dislikes apply negative delta
• Optimistic UI: Comments appear instantly with a "Sending..." state, then update with server response

7. Country Statistics Dashboard

/stats/country/{name}/dashboard at backend/src/handlers/stats.rs:184 returns a comprehensive response with:

• National averages (temp, humidity, wind, UV index)
• Monthly temperature trends (line chart)
• Historic yearly extremes (min/max bar chart)
• Yearly evolution
• Hottest and coldest cities today (top 5)
• In-memory climate alerts (heat wave, cold wave, strong wind, extreme humidity, drought, climate anomaly)
• Database alerts from get_country_alerts()

8. Weather Clustering

get_country_city_clusters() at backend/migrations/014_add_bulk_predictions_and_rankings.sql classifies cities within a country into 8 clusters: warm/humid/windy · warm/humid/calm · warm/dry/windy · warm/dry/calm · cool/humid/windy · cool/humid/calm · cool/dry/windy · cool/dry/calm

Each city gets a similarity score relative to the national average, and clusters show their size.

9. City Leaderboards & Forecast Scoreboard

• City Forecast Leaderboard: Ranks cities by a composite score (ideal temp = 22°C, penalties for wind, humidity deviation, alert count)
• Forecast Scoreboard: Ranks forecasts by weighted accuracy (reputation-weighted), comment count, reputation score, and vote balance

10. Volatility Risk Assessment

proc_classify_city_risk() classifies cities based on 7-day temperature volatility:

• STABLE CLIMATE (diff ≤ 10°C)
• MODERATE RISK (diff 10-20°C)
• EXTREME VOLATILITY (diff > 20°C)

11. Anomaly Detection

proc_detect_city_anomaly() compares the latest temperature against the historical average. Flagged as anomaly if deviation exceeds 10°C. The frontend visualizes this with a bar chart showing the exact deviation.

12. City Trust Score

proc_audit_city_trust() computes average user accuracy rating per city and labels it:

• HIGH TRUST (> 4.0) — "Data validated by users"
• STABLE (2.5-4.0)
• LOW TRUST (< 2.5) — "Check sensors"

13. Seasonal Comparisons

get_city_seasonal_comparison() at backend/migrations/014_add_bulk_predictions_and_rankings.sql compares current conditions against:

• Same day of month across all prior years
• Entire seasonal baseline (winter/spring/summer/autumn)
• Delta score aggregates temperature, wind, and humidity differences

14. User Power Score

/stats/user/{id}/power computes a user's influence score as: reputation × 0.5 + comments × 5 × 0.3 + reactions × 2 × 0.2.

15. Interactive Map

MapView.tsx at frontend/src/components/MapView.tsx uses Leaflet with:

• Dynamic city loading based on map bounds (zoom level 6+)
• City markers with weather popups (temp, wind, humidity)
• Favorite toggle directly on map markers
• "You must be logged in" tooltip for unauthenticated favorite attempts
• Fly-to animation when searching cities

16. Token Refresh with Request Queue

The Axios interceptor at frontend/src/services/api.ts handles 401 errors by:

• Queuing all pending requests while refreshing the token
• Processing the queue atomically after successful refresh
• Redirecting to login on refresh failure

17. Optimistic Comment UI

Comments submitted on CityDetails page appear immediately with a "Sending..." indicator and transition to the server response seamlessly. On failure, they roll back and restore the input.

18. Seed Script with Resume Support

seed_meteo.py at scripts/seed_meteo.py intelligently:

• Selects 40 most populous cities per country
• Fetches 853 days of historical data from Open-Meteo Archive API
• Handles 429 rate limits with 30-minute backoff and countdown
• Saves progress to progress.txt for resumable seeding
• Runs update_country() at the end to normalize country associations

19. Alert Generation Script

generate_alerts.py creates extreme weather forecasts to trigger the alert system — useful for testing.

20. Documentation in Romanian

See documentation/referat_prognoza_meteo.docx for the full academic paper in Romanian covering architecture, algorithms, and methodology.

---

How to Run

Prerequisites

• PostgreSQL (running)
• Rust (latest stable)
• Node.js 20+
• Python 3.10+

Setup

1. Backend environment — create backend/.env:

   DATABASE_URL=postgres://user:pass@localhost/prognoza_meteo
   JWT_SECRET=your-secret-key
   GOOGLE_CLIENT_ID=your-google-client-id
   GOOGLE_CLIENT_SECRET=your-google-client-secret
   

2. Start backend (migrations run automatically):

   cd backend
   cargo run
   # Server starts at http://localhost:3000

3. Start frontend (in a separate terminal):

   cd frontend
   npm install    # first time only
   npm run dev
   # Opens at http://localhost:5173

4. Seed the database (populates 600 cities × 853 days ≈ 5M rows):

   cd scripts
   python seed_meteo.py

   The script is resumable — if interrupted, it continues from where it left off.

5. (Optional) Generate test alerts:

   cd scripts
   python generate_alerts.py
   python generate_romania_anomaly.py