Tech-Rozvi/demo-drought

# RozviDrought Demo Inference System ## Overview This repository demonstrates how to run drought inference using the **rozvidrought ecosystem**. The system converts spatial inputs into drought predictions using packaged models. Pipeline: Geometry → Data extraction → Feature engineering → Subsystem models → Fusion model → Drought classification This repository is a **reference implementation** showing how to connect: - rozvidrought-datasets - rozvidrought-inputs - rozvidrought-subsystems - rozvidrought into a working API and frontend. --- ## Important Rule **Do not start from the repository root.** Always start inside: RozviDrought/ All runtime code lives there. --- ## Repository Structure RozviDrought/ │ ├── app/ │ ├── api/ │ ├── services/ │ ├── schemas/ │ └── frontend/ │ ├── provenance/ ├── runs/ ├── tests/ ├── docs/ │ ├── run_api.py ├── requirements.txt └── README.md --- ## System Requirements Python: Python 3.10 or newer Operating system: Windows, Linux, or macOS --- ## Installation Clone the repository. git clone github.com Move into the application folder. cd demo-drought cd RozviDrought Install dependencies. pip install -r requirements.txt --- ## Required Data Large datasets are **not stored in Git**. You must obtain them separately. --- ## Required Dataset The system requires one serving dataset: master_inputs_long_198001_205012.parquet --- ## Where to Place the Parquet File Place the dataset here: RozviDrought/data/master_inputs/ Example: RozviDrought/ data/ master_inputs/ master_inputs_long_198001_205012.parquet This file is the **serving dataset** used during inference. The API reads this file directly. --- ## What the Dataset Contains The dataset contains: pixel_id row col lon lat scenario yyyymm t2m d2m pet sm ndvi tws Each row represents: one pixel one month --- ## Start the API From inside: RozviDrought/ Run: python run_api.py …