social-data-attribution-w2/scripts/README.md

# Script Layout

I keep `scripts/` grouped by purpose instead of leaving unrelated entrypoints at the root.

- `scripts/bootstrap/`: local environment setup and credential wrappers
- `scripts/docs/`: document and PDF extraction helpers
- `scripts/enrichment/`: enrichment workers and related debug or upload helpers
- `scripts/manifests/`: manifest generation and pre-splitting utilities
- `scripts/query_data/`: query-dataset conversion, download, scoring, and upload tools
- `scripts/sampling/`: stratified sampling extraction and bin-splitting helpers
- `scripts/transform/`: general data reshaping and file-format conversion utilities
- `scripts/validation/`: reporting, coverage checks, and validation entrypoints
- `scripts/modal/`: Modal execution paths
- `scripts/slurm/`: Slurm launchers and cluster wrappers
- `scripts/analysis/`: existing analysis-oriented helpers retained in place for now

When adding a new script, prefer the most specific folder above instead of the `scripts/` root.