Every dataset published on benthic.io, and where its documentation lives. Six datasets across two collections; both numbers grow as datasets are added.
Each row is checked by scripts/check_docs.py in CI: the reference page, the OpenAPI spec, the signed manifest, and the pipeline README must all exist. If a dataset is added without documentation, the build fails.
NGOpen
Federal spending, nonprofit, and legislative data, published as five connected PostgREST APIs over PostgreSQL/PostGIS.
Signed collection manifest: ngopen/collection.json
Parts
Vehicle, component, and identifier reference datasets. NHTSA/vPIC is the first member and is deliberately not an NGOpen dataset.
Signed collection manifest: parts/collection.json
| Dataset | API | OpenAPI | Manifest | Pipeline |
|---|---|---|---|---|
| NHTSA vPIC Vehicle Information | /parts/NHTSA/ |
swagger | bdp | README |
Provenance
Read from the signed manifests at build time. migrated means the serving database was built by the commit named here and verified against the previous database by a structural comparison.
| Dataset | Status | Migrated | Commit | Relations | Columns | Queryable | Updated |
|---|---|---|---|---|---|---|---|
usaspending |
migrated | 2026-09-21 | b25eba8404 |
86 | 3621 | 66 | Monthly |
irs_ng |
migrated | 2026-08-14 | 66f585560f |
18 | 388 | 18 | Monthly |
samer |
migrated | 2026-08-11 | 66f585560f |
2 | 53 | 2 | Monthly |
up_cdmaps |
migrated | 2026-08-11 | 66f585560f |
1 | 21 | 1 | Static — run once |
usp_cl |
migrated | 2026-08-11 | 66f585560f |
12 | 172 | 12 | Weekly |
nhtsa |
migrated | 2026-09-24 | 81e1d7de3c |
107 | 383 | 106 | Monthly |
Reading order
If you are new to the data, read the APIs overview first — it has the join paths and worked cross-dataset queries, and it is the only page that explains how the collections fit together.
If you are verifying a dataset, read the BDP specification — it explains what the signatures in the manifest column above actually prove.
If you are running a pipeline, start at the pipeline README. Each documents that dataset’s source, its real per-stage behaviour, and the things that will bite you — which is more than the reference page can tell you.