docs/data_source_registry.md

AutoFarm Data Source Registry

This document records the public sources used by the active data bootstrap pipeline and the local cache that supports reproducible rebuilds.

Provenance Status

• exact: the upstream endpoint or asset is known and documented.
• page-only: the upstream source page is known, but the exact downloaded file URL is not preserved locally.
• repo-derived: the local file is derived from another source already present in the repository.
• unresolved: the exact upstream acquisition path is not recoverable from the current repository state.

Assets Used By The Current Pipeline

Local asset	Current use	Upstream source	Provenance status	Evidence
data_local/downloads/usda_soil/	zone_state_bootstrap cache	USDA Soil Data Access POST endpoint: https://sdmdataaccess.sc.egov.usda.gov/Tabular/post.rest	exact	Used by build_zone_state_bootstrap.py.
live Open-Meteo archive queries	zone_state_bootstrap	Archive API: https://archive-api.open-meteo.com/v1/archive	exact	Used by build_zone_state_bootstrap.py.
live Open-Meteo forecast queries	zone_state_bootstrap	Forecast API: https://api.open-meteo.com/v1/forecast	exact	Used by build_zone_state_bootstrap.py.
live Open-Meteo elevation queries	zone_state_bootstrap	Elevation API: https://api.open-meteo.com/v1/elevation	exact	Used by build_zone_state_bootstrap.py.
live SoilGrids fallback queries	zone_state_bootstrap fallback	SoilGrids REST query endpoint: https://rest.isric.org/soilgrids/v2.0/properties/query	exact	Written into the dataset card by build_zone_state_bootstrap.py.

Rebuild Notes

When recreating the processed public-data outputs in a fresh environment:

1. run python scripts/run_public_data_pipeline.py,
2. verify that data/processed/zone_state_bootstrap.parquet exists,
3. confirm that data/processed/zone_state_bootstrap.dataset_card.json records the active upstream sources.