zerisinyu/ADI-Water-Dashboard
Domaine:
environment and energydigital infrastructureType de record:
softwareproject
Créateur:
zer
HĂ´te:
A data science dashboard project for monitoring water utilities in African countries
# đź’§ Water Utility Performance Dashboard
A role-based analytics platform for water & sanitation utilities — tracking access & coverage, service quality, financial health, and production across countries, with AI-assisted insights and forecasting.
**▶ Try the live demo** — sign in with `admin` / `admin123`; The first load takes a minute or two while packages install.
---
## Overview
The dashboard turns raw utility CSVs into a single, decision-oriented view of how a water and sanitation utility is performing. It combines secure role-based access, a DuckDB analytical pipeline, and an optional bring-your-own-key AI assistant, at national, city, and zone levels.
It was proudly built as a capstone project with Athena Infonomics and the Africa Utility Data Collaborative (AUDC), by **Fusion 4** — a team from the Applied Data Institute '25 cohort at Equitech Futures.
## ✨ Highlights
- **Five domain pages** — Access & Coverage, Service Quality, Financial Health, Production, and an Executive home page that rolls all four into one headline metric each, plus current risks and wins.
- **Insights & Forecasting** — Holt-Winters time-series forecasts with confidence bands, seasonal decomposition, what-if scenario modelling, and a metric-correlation explorer.
- **MajiBot AI assistant (BYOK)** — an optional chat assistant grounded in the on-screen data, plus AI-drafted board briefs and to-do lists. Supports Gemini, GLM, Grok, OpenAI, DeepSeek, and OpenRouter; degrades gracefully to rule-based output when no key is set.
- **Role-based access** — master, country-admin, analyst, and viewer roles, each scoped to the data they're allowed to see.
- **Cross-country / cross-zone benchmarking** — a peer-comparison radar that narrows from country to zone once a single country is selected.
- **Honest about gaps** — measures that aren't in the data are shown as explicit data-gap panels rather than estimated.
- **In-app documentation** — a built-in reference page exp …