Abson-dev/kenya-fnp-db

# Kenya Soil / Food / Nutrition / Policy database A configuration-driven, provenance-tracked acquisition pipeline that assembles the Kenya dataset bundle into a single layered database keyed on a master county and sub-county crosswalk. It is built for county-level profiling and for the soil-to-nutrition analysis at the centre of the bundle. Author: Aboubacar HEMA --- ## Table of contents 1. Design in one paragraph 2. Repository layout 3. The `kenyadb` package 4. The `data` directory 5. Quick start and CLI 6. Pipeline stages 7. Data inventory by layer 8. Acquisition status 9. Database schemas and tables 10. Analysis layer 11. Validation 12. Provenance and reproducibility 13. Extending the pipeline 14. Requirements 15. License and citation 16. Caveats carried from the bundle --- ## Design in one paragraph The bundle is treated as a layered system, not one monolithic table. A master county and sub-county crosswalk (from COD-AB boundaries, enriched by the 2019 census) is the spine, and every later indicator is appended to that crosswalk rather than merged table-to-table. Soil, food, health and policy each form a thematic schema. Heavy artefacts (rasters, survey microdata, strategy PDFs) stay on disk as files; the database holds the tidy indicators, the crosswalk, the flattened source registry, and a full provenance ledger that records publisher, mirror, licence, checksum and extraction date for every object. This mirrors the integration rules in the bundle document: three parallel keys (spatial, denominator, provenance) and a strict separation between gridded predictions, legacy polygons, measured points, survey microdata and policy text. --- ## Repository layout ``` kenya_fnp_db/ config/ sources.yaml single source of truth: 30 sources across 5 layers plus remote sensing run_all.py main CLI: acquire -> extract -> crosswalk -> transform -> build build_action_plan.py standalone runner for the Action Plan extractor build_census. …