registry¶

Metadata registry driving frontend UI copy, the model picker, and the AI interpreter's prompt context. See Concepts § The registry.

registry ¶

USE_CASE_REGISTRY `module-attribute` ¶

USE_CASE_REGISTRY: dict[str, UseCaseInfo] = {'flood': UseCaseInfo(id='flood', name='Flood Risk Monitoring & Mitigation', icon='🌊', description='This tool analyses terrain, rainfall patterns, and satellite radar imagery to identify which areas are at risk of flooding and how severe that flooding is likely to be. It produces a risk map showing low, medium, high, and very high risk zones.', climate_importance='Floods are the most frequent and costly natural disaster in Africa, affecting millions of people and destroying crops, infrastructure, and livelihoods. Early identification of flood-prone zones enables governments and communities to invest in resilient infrastructure, plan evacuations, and protect the most vulnerable.', dependent_variable='Flood risk class (Low / Medium / High / Very High)', independent_variables=['Elevation (how high the land is)', 'Terrain slope (how steep the hillside is)', 'Flow accumulation (how much water drains into an area)', 'Topographic Wetness Index (how waterlogged land tends to get)', 'Distance to nearest river', '3-day and 7-day cumulative rainfall', 'Satellite SAR radar backscatter (detects surface water)', 'Land cover type (e.g., forest, bare soil, urban)'], best_model='Random Forest Classifier', model_accuracy='F1-score: 0.87 on African flood inventory benchmarks', output_type='4-class risk map (GeoTIFF + interactive map) + SHAP feature chart', date_guidance='Select a rainy season period for best results (e.g., March–May or Oct–Dec)', min_months=1, max_years=5, model_options=[ModelOption('rf', 'Random Forest', True, 'Fastest · most interpretable'), ModelOption('xgboost', 'XGBoost', False, 'Higher accuracy · longer run'), ModelOption('ensemble', 'Ensemble (RF + XGBoost)', False, 'Best overall · slowest')], default_model='rf', single_area_only=True), 'food_security': UseCaseInfo(id='food_security', name='Food Security Assessment', icon='🌾', description='This tool predicts crop yields and identifies areas at risk of food shortages by analysing vegetation health, rainfall, temperature, and soil moisture using satellite data. It shows where harvests are likely to be poor and flags food-insecure zones.', climate_importance='Over 250 million Africans are food insecure. Climate variability — particularly erratic rainfall and drought — is the leading driver. Early warning of poor harvest seasons enables governments, NGOs, and farmers to mobilise resources, adjust planting decisions, and prevent famine.', dependent_variable='Crop yield (tonnes/ha) or food insecurity risk class', independent_variables=['NDVI — vegetation greenness during growing season', 'EVI — enhanced vegetation index', 'Land surface temperature', 'Seasonal cumulative rainfall', 'Rainfall anomaly vs. long-term average', 'Soil moisture', 'Vegetation Condition Index (VCI)', 'Vegetation Health Index (VHI)', 'Standardised Precipitation Index (SPI-3)', 'Growing season length'], best_model='Random Forest + XGBoost + mean-probability ensemble', model_accuracy='R² = 0.82 against FAO yield records', output_type='Yield prediction raster + food insecurity map + NDVI trend chart', date_guidance='Select a full growing season (at least 3 months). Long rains: Mar–May. Short rains: Oct–Dec.', min_months=3, max_years=10, model_options=[ModelOption('rf', 'Random Forest', True, 'Fast · interpretable'), ModelOption('xgboost', 'XGBoost', False, 'High accuracy · longer run'), ModelOption('ensemble', 'Ensemble (RF + XGBoost)', False, 'Best overall · slower')], default_model='rf', single_area_only=False), 'disease': UseCaseInfo(id='disease', name='Climate-Driven Disease Surveillance', icon='🦟', description='This tool maps where climate conditions — warmth, rainfall, and standing water — create the ideal environment for disease outbreaks like malaria and cholera. It shows current risk zones and provides a 4-week outbreak probability forecast.', climate_importance='Vector-borne and water-borne diseases are highly sensitive to climate. Warming temperatures and changing rainfall expand the geographic range of mosquitoes and contaminate water sources. Spatial early warning systems help health ministries pre-position medicines and response teams.', dependent_variable='Disease outbreak risk class (Low / Medium / High) or weekly case count forecast', independent_variables=['4-week cumulative rainfall', 'Mean and maximum temperature', 'Relative humidity', 'Standing water surface extent (from satellite)', 'Population density', 'Distance to nearest health facility', 'Elevation', 'Historical lagged case counts (if available)'], best_model='Gradient Boosting + XGBoost + mean-probability ensemble', model_accuracy='AUC: 0.84 on East African malaria benchmarks', output_type='Weekly risk map + hotspot polygons + 4-week forecast chart', date_guidance='Select the most recent 3-6 months for current risk; up to 2 years for trend analysis.', min_months=3, max_years=5, model_options=[ModelOption('gbm', 'Gradient Boosting', True, 'Default disease model'), ModelOption('xgboost', 'XGBoost', False, 'High accuracy · longer run'), ModelOption('ensemble', 'Ensemble (GBM + XGBoost)', False, 'Best overall · mean softmax')], default_model='gbm', single_area_only=False), 'land_degradation': UseCaseInfo(id='land_degradation', name='Rangeland Dynamics & Land Degradation', icon='🏜️', description='This tool detects where land is losing its ability to support vegetation and livestock by analysing satellite imagery over time. It identifies degraded areas, measures how severe the degradation is, and shows when the decline started.', climate_importance='Land degradation affects 65% of African land and is accelerated by climate change, overgrazing, and deforestation. It reduces agricultural productivity, triggers conflict over resources, and contributes to CO₂ emissions. Early detection enables targeted restoration investments.', dependent_variable='Land degradation severity score (0–100) and change class', independent_variables=['NDVI trend slope over the analysis period', 'Year and magnitude of detected vegetation breakpoint', 'Bare soil fraction', 'Bare Soil Index (BSI)', 'Soil organic carbon content', 'Rainfall erosivity (RUSLE R-factor)', 'Land cover change indicator', 'Terrain slope', 'Livestock density'], best_model='Random Forest + LightGBM (change detection) · majority-vote ensemble', model_accuracy='Overall accuracy: 89% on LUCAS land cover benchmarks', output_type='Annual change map + degradation severity raster + NDVI trend chart', date_guidance='Select at least 5 years for meaningful trend detection; 10+ years recommended.', min_months=60, max_years=25, model_options=[ModelOption('rf', 'Random Forest', False, 'Fast change detection'), ModelOption('lgbm', 'LightGBM', True, 'Best land baseline'), ModelOption('ensemble', 'Ensemble (RF + LightGBM)', False, 'Best overall · majority vote')], default_model='lgbm', single_area_only=False), 'drought': UseCaseInfo(id='drought', name='Drought Monitoring & Early Warning', icon='☀️', description='This tool measures current drought conditions and forecasts whether drought is likely to worsen using a Composite Drought Index (CDI) that combines rainfall deficits, vegetation stress, and temperature extremes. It shows drought severity by area and how conditions have changed over time.', climate_importance="Drought is Africa's most damaging climate hazard, affecting over 40 million people and causing billions in agricultural losses annually. Multi-index drought monitoring enables governments to trigger contingency plans, redirect water resources, and support pastoralists and farmers before crisis deepens.", dependent_variable='Composite Drought Index (CDI) class (Extreme / Severe / Moderate / Mild / Near Normal / Wet)', independent_variables=['Precipitation Drought Index (PDI) — rainfall anomaly from ERA5', 'Temperature Drought Index (TDI) — heat stress from ERA5', 'Vegetation Drought Index (VDI) — vegetation stress from MODIS NDVI', 'Standardised Precipitation Index (SPI) at 1, 3, 6, 12 months', 'Standardised Precipitation-Evapotranspiration Index (SPEI)', 'Vegetation Condition Index (VCI)', 'Vegetation Health Index (VHI)', 'Soil moisture anomaly', 'Consecutive dry days'], best_model='drought-monitoring CDI ensemble (PDI + TDI + VDI) + LSTM forecast', model_accuracy='CDI validated against PDSI benchmarks; r=0.91 on African drought events', output_type='Annual CDI maps + CDI sub-index time-series chart + 6-month LSTM forecast', date_guidance='Select 10–30 years for climatological baseline. Minimum 5 years for trend detection.', min_months=60, max_years=30, model_options=[ModelOption('lstm', 'LSTM Forecast', True, 'MC Dropout uncertainty · 6-month horizon'), ModelOption('drought_monitoring', 'drought-monitoring statistical', False, 'Package built-in · faster')], default_model='lstm', single_area_only=True)}

ModelOption `dataclass` ¶

ModelOption(id: str, label: str, recommended: bool, note: str)

One selectable model for a use case, as shown to a frontend user.

id is the value accepted by extra_params["model_type"] in run_analysis.

UseCaseInfo `dataclass` ¶

UseCaseInfo(id: str, name: str, icon: str, description: str, climate_importance: str, dependent_variable: str, independent_variables: list[str], best_model: str, model_accuracy: str, output_type: str, date_guidance: str, min_months: int, max_years: int, model_options: list[ModelOption], default_model: str, single_area_only: bool)

Display and configuration metadata for one domain module (e.g. "disease").

Drives frontend UI copy (name, icon, description), the AI interpreter's prompt context (climate_importance, dependent/independent variables), and the model picker (model_options, default_model). Populated once per module in USE_CASE_REGISTRY and looked up via get_use_case_info.

get_use_case_info ¶

get_use_case_info(module_id: str) -> UseCaseInfo

Look up registry metadata for a module id (e.g. "flood", "disease").

Raises KeyError listing the valid ids if module_id is not registered.

Source code in registry.py

def get_use_case_info(module_id: str) -> UseCaseInfo:
    """Look up registry metadata for a module id (e.g. "flood", "disease").

    Raises `KeyError` listing the valid ids if `module_id` is not registered.
    """
    if module_id not in USE_CASE_REGISTRY:
        raise KeyError(f"Unknown module: '{module_id}'. Available: {sorted(USE_CASE_REGISTRY)}")
    return USE_CASE_REGISTRY[module_id]

registry¶

registry ¶

USE_CASE_REGISTRY module-attribute ¶

ModelOption dataclass ¶

UseCaseInfo dataclass ¶

get_use_case_info ¶

USE_CASE_REGISTRY `module-attribute` ¶

ModelOption `dataclass` ¶

UseCaseInfo `dataclass` ¶