# Database releases and package compatibility This page is generated by `scripts/kg_compat_page.py` from the release snapshots committed under `database/releases/` and must not be edited by hand; regenerate it with `python scripts/kg_compat_page.py` after a release is stamped, and `--check` fails when it is stale. ## Pairs A knowledge-graph release and a package version are used as a pair. The client refuses to read a release whose paired package is not the installed one unless the installed version also reads it (listed under *reads it*), so install the package named here for the release you query. | KG release | KG version | paired package | reads it | | --- | --- | --- | --- | | 2026.09.21-ab6d8c5d | 1.2 | v1.2.1 | v1.2.1, v1.3.0, v1.4.0, v1.4.1, v1.5.0, v1.6.0, v1.6.1 | | 2026.10.02-833970cd | 2.0 | v1.6.1 | v1.2.1, v1.3.0, v1.4.0, v1.4.1, v1.5.0, v1.6.0, v1.6.1 | | 2026.10.06-4b293d34 | 3.0 | v1.6.2 | v1.6.2 | ## Releases | KG release | KG version | built | commit | package version at build | tag | | --- | --- | --- | --- | --- | --- | | 2026.09.21-ab6d8c5d | 1.2 | 2026-09-21T20:25:23.207396+00:00 | 7715ee35 | 1.2.1 | v1.2.1 | | 2026.10.02-833970cd | 2.0 | 2026-10-02T05:55:48-05:00 | 833970cd | 1.6.1 | v1.6.1 | | 2026.10.06-4b293d34 | 3.0 | 2026-10-06T04:42:48-05:00 | 4b293d34 | 1.6.2 | v1.6.2 | ## Compatibility | package tag | 2026.09.21-ab6d8c5d | 2026.10.02-833970cd | 2026.10.06-4b293d34 | | --- | --- | --- | --- | | v1.2.0 | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | | v1.2.1 | compatible | compatible | incompatible (20 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenHoepfner2014Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, HetHillenmeyer2008Dataset, HomHillenmeyer2008Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | | v1.3.0 | compatible | compatible | incompatible (20 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenHoepfner2014Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, HetHillenmeyer2008Dataset, HomHillenmeyer2008Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | | v1.4.0 | compatible | compatible | incompatible (20 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenHoepfner2014Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, HetHillenmeyer2008Dataset, HomHillenmeyer2008Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | | v1.4.1 | compatible | compatible | incompatible (20 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenHoepfner2014Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, HetHillenmeyer2008Dataset, HomHillenmeyer2008Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | | v1.5.0 | compatible | compatible | incompatible (20 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenHoepfner2014Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, HetHillenmeyer2008Dataset, HomHillenmeyer2008Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | | v1.6.0 | compatible | compatible | incompatible (20 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenHoepfner2014Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, HetHillenmeyer2008Dataset, HomHillenmeyer2008Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | | v1.6.1 | compatible | compatible | incompatible (20 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenHoepfner2014Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, HetHillenmeyer2008Dataset, HomHillenmeyer2008Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | | v1.6.2 | incompatible (17 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | incompatible (17 of 51 datasets drift: DmfCostanzo2016Dataset, DmfKuzmin2018Dataset, DmfKuzmin2020Dataset, DmiCostanzo2016Dataset, DmiKuzmin2018Dataset, DmiKuzmin2020Dataset, EnvChemgenVanacloig2022Dataset, EnvChemgenWildenhain2015Dataset, SmfBaryshnikova2010Dataset, SmfCostanzo2016Dataset, SmfKuzmin2018Dataset, SmfKuzmin2020Dataset, SmfODuibhir2014Dataset, TmfKuzmin2018Dataset, TmfKuzmin2020Dataset, TmiKuzmin2018Dataset, TmiKuzmin2020Dataset) | compatible | | v1.7.0 | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | | v1.7.1 | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | | v1.7.2 | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | incompatible (all 51 datasets drift) | ## How compatibility is decided Every served record is serialized under the schema contract of the commit that built it, and the release snapshot records, per dataset, the contract fingerprints of that dataset's schema closure. A package tag is compatible with a release when the schema surface at that tag (`torchcell/datamodels/schema.py` and `pydant.py`) reproduces every one of those fingerprints, which is the same drift check the admission gate runs before an incremental import, turned around to face a client. An incompatible verdict names the datasets whose records that package version would serialize differently, and the client refuses the whole release at connect (`releases.require_paired`) rather than reading the rest.