SKY CAST — Enterprise Climate Intelligence Platform

From generic weather apps to enterprise climate intelligence. Official AEMET data + crowdsourcing (Waze model) + smart alerts + REST API.

---

Elevator Pitch

Problem: Generic weather apps show "23°C and sunny" — but a city council needs to know if tomorrow's wind will exceed 50 km/h to close parks, and a logistics company needs to decide whether to delay fleets due to heavy rain on a specific route. No solution combines official data with collaborative intelligence for enterprise decision-making.

Hypothesis: Combining official AEMET data with geo-validated crowdsourcing (Waze model) and configurable alert thresholds can create a climate intelligence system that outperforms generic apps and delivers real value to councils, logistics companies, and enterprises.

Solution: SkyCast — a platform that evolved through 4 phases (from Streamlit + CSV prototype to production-grade FastAPI + PostgreSQL + Docker + CI/CD), integrating AEMET OpenData, collaborative reports, smart alerts, documented REST API, and 158 automated tests ensuring production quality.

---

Problem

Generic weather apps don't cover specific business needs:

• A city council needs to know if tomorrow's wind > 50 km/h to close parks, not just "23°C and sunny".
• A logistics company needs to decide whether to delay fleets due to heavy rain on a specific route.
• A municipal technician cross-references official sensor data with citizen reports to decide whether to activate a frost protocol.

SkyCast solves this by combining official data (AEMET OpenData), crowdsourcing (geo-validated manual reports), and analytics (alerts, anomalies, historical trends).

Key Metrics

Metric	Value
Automated Tests	158 (33 files, 0 failures)
API Endpoints	10+ (REST + Swagger)
Data Sources	AEMET OpenData + OpenWeatherMap (fallback) + Crowdsourcing
Refresh Rate	Every 2 hours (APScheduler)
Security	JWT + SHA-256 + salt + rate limiting
Architecture Layers	4 (API, Dashboard, ETL, Scheduling)
Evolution Phases	4 (V1 protoype -> Production)
Database Models	9 (SQLAlchemy ORM)
Test Coverage	Auth, Alerts, Validators, ETL, Anomaly Detection, Cache, Lineage

Differentiator

Factor	Generic Weather Apps	SkyCast
Data Source	Private API	AEMET OpenData (official Spain)
Crowdsourcing	No	Geo-validated manual reports
Alerts	Generic	User-configurable thresholds
Historical Data	Limited	Full CRUD + ETL with lineage
Third-party API	No	Documented REST (Swagger)
Audience	Consumer	Businesses, councils, logistics

Architecture

FastAPI + Pydantic -- REST API --+
                                  +-- PostgreSQL (SQLAlchemy ORM)
Streamlit + Plotly --- Dashboard -+
                                  |
APScheduler --- AEMET API --------+
    (every 2h, fallback OpenWeatherMap)

ETL Flow (full lineage)

Extract (AEMET JSON) -> Transform (Pandas: nulls, duplicates, types)
                      -> Load (SQLAlchemy) -> LineageLogger (trace)

Each step logs input/output/discarded rows with LineageLogger.

Tech Stack

Layer	Technology
API	FastAPI + Uvicorn + Pydantic + async
Database	PostgreSQL 16 / SQLite + SQLAlchemy ORM (9 models)
Dashboard	Streamlit + Plotly + Geopandas + Folium
ETL	Pandas + NumPy + LineageLogger
Auth	JWT + SHA-256 + unique salt
Cache	Redis (optional, in-memory fallback)
Scheduling	APScheduler (AEMET fetch every 2h)
Containers	Docker + Docker Compose
CI/CD	GitHub Actions (pytest + ruff + Docker build)
Testing	158 tests, pytest + httpx, pre-push hooks

Quick Start

git clone https://github.com/juandelaf1/SkyCast.git
cd SkyCast
python -m venv .venv
pip install -r requirements.txt
cp .env.example .env   # Add AEMET_API_KEY
python -m app.db       # Init DB
uvicorn app.main:app --reload --port 8000
streamlit run app/dashboard/app.py --server.port 8501

Docs: http://localhost:8000/docs | Dashboard: http://localhost:8501

API Endpoints

Method	Endpoint	Description
POST	/api/v1/auth/register	Register
POST	/api/v1/auth/login	JWT Login
GET	/api/v1/clima?lat=&lon=	Current weather
GET	/api/v1/geo/{city}	Geocoding
GET	/api/v1/registros?page=&limit=	Paginated history
POST	/api/v1/registros	Manual report (crowdsourcing)
POST	/api/v1/comparar	Manual vs AEMET
GET/POST	/api/v1/alertas	Configurable thresholds

Security

• Passwords: SHA-256 + salt (16 bytes) per user
• Tokens: JWT HS256, 24h expiration
• Validation: Pydantic + physical ranges (-20/60°C, 0-100% humidity)
• Rate limiting: slowapi (30 req/min climate, 10 geo)
• Pre-push hooks: lint + tests + anti-leak secrets

Quality (158 tests, 0 failures, CI/CD)

pytest --cov=app --cov-report=html   # 158 tests, 33 files
ruff check app/ tests/               # 0 errors

Coverage: API endpoints, auth, alerts, validators, ETL (extract/transform/load), haversine, anomaly detection, cache, lineage logger, external services mocked.

Vision: Logistics Platform

SkyCast is designed as the climate module of a larger logistics platform. REST endpoints allow routing, fleet, and insurance systems to consume historical and real-time climate data without coupling. Natural next steps:

1. Push alerts (email/Telegram) for personalized thresholds
2. Geo-fences: trigger alerts when locations exceed thresholds
3. Routing API integration (calculate weather-related delays)
4. Predictive model (temp/humidity regression to 48h)

---

Project Evolution

Phase	Project	Stack	Milestone
F1	SkyCast V1	Streamlit + CSV	Functional prototype
F2	ClimApp	Flask MVC + AEMET	Layered architecture, 66 tests
F3	Vortex	FastAPI/Flask + PostgreSQL	ETL, lineage, traceability
F4-Pre	SkyCast V2 Pre	FastAPI + Docker	JWT with salt, anomalies
F4	SkyCast	FastAPI + PostgreSQL + Docker	158 tests, CI/CD, production-ready

---

Author

Juan de la Fuente — @juandelaf1 · juandelafuentelarrocca@gmail.com

MIT © 2026