How should we represent the schema when there is a growing number of things to support

[borgr]

We suggested the option to have subparts (per example, per benchmark, per evaluarion...) that have links between them and there are combinations of them that include different parts (e.g., agents might add another component and use the per model data multiple times)
@Erotemic suggested stack and a format like json-ld (json schemas that link to each other).

[Erotemic]

Small correction, it is spelled STAC. Reference: https://stacspec.org/en/about/stac-spec/

[borgr]

No embarrassed emoji in github...