abdeladimyacine-create/goutte_eau_mvp

How abdeladimyacine-create/goutte_eau_mvp works

Goutte d’Eau is a Paris rainfall-risk forecasting MVP. Its main workflow collects Infoclimat observations, cleans and aggregates them into daily records, trains a next-day model, then serves predictions through a REST API that the Streamlit interface calls for an end user. The graph also includes station selection and the separate raw and daily database stages. The README documents this full chain; the sampled source confirms the collection, transformation, training, API, and interface calls. Other unsampled details are represented only where documented.

Main entry points: api.py Forecast API · entrainement.py Model training · meteo.db Daily observations · collecte.py Observation collection

Components 12

Persistent state

Raw observations data/meteo.db
Daily observations data/meteo.db
Model and metrics model/modele.joblib

Other components

End user
Infoclimat API

Connections 13

Station selection → Infoclimat API: compares stations choisir_station.py
Observation collection → Infoclimat API: fetches observations collecte.py
Observation collection → Raw observations: writes raw data collecte.py
Daily aggregation → Raw observations: reads raw data nettoyage.py
Daily aggregation → Daily observations: writes daily data nettoyage.py
Model training → Daily observations: reads daily data entrainement.py
Model training → Feature engineering: prepares features entrainement.py
Model training → Model and metrics: saves model and metrics README.md
Forecast API → Daily observations: reads daily data api.py
Forecast API → Feature engineering: prepares prediction features api.py
Forecast API → Model and metrics: loads model and metrics README.md
Demo interface → Forecast API: requests forecasts app.py
End user → Demo interface: chooses forecast date app.py

Drawn by GitDiagram on November 14, 2026 from the repository's file tree, README and source files. Read as Markdown ·