Dataset & Metadata Sections

Dataset node (./)

Field	Required	Notes
name	✅	Appears as the card title. Keep it short and action-oriented.
description	✅	Full markdown-free text. The first 240 characters are shown on the card, the rest lives in the detail modal/JSON.
serviceType	✅	synchronous or asynchronous. Values outside this set will hide filter options.
URL	✅	Destination for users (repository, API explorer, documentation).
license	✅	Either a single @id to a remote resource or an inline object with name + url.
author	✅	Person or organisation. Provide name and url for attribution.
datePublished	✅	ISO-8601 date. Update whenever you make a significant change.
memoryRequirements	⚙️	Free-form string describing CPU/GPU requirements.
processorRequirements	⚙️	Array of CPU/GPU strings when multiple SKUs are supported.
softwareRequirements	⚙️	Array of references (objects or @ids) pointing to nodes that describe dependencies.
logo	⚙️	Reference to icon.png (or similar). Must exist on disk.

License node

Define once inside @graph when you need extra fields:

{
  "@id": "https://opensource.org/licenses/MIT",
  "@type": "CreativeWork",
  "name": "MIT License"
}

Author node

You can describe a person or organisation. Both can include contact URLs.

{
  "@id": "#maintainer",
  "@type": "Organization",
  "name": "GRyCAP",
  "url": "https://www.grycap.upv.es"
}

Software requirements

Represent each dependency with its own node:

{
  "@id": "#torch",
  "@type": "SoftwareApplication",
  "name": "PyTorch",
  "version": "2.1",
  "url": "https://pytorch.org"
}

Mention the node inside the dataset softwareRequirements array:

"softwareRequirements": [{ "@id": "#torch" }, { "@id": "#ultralytics" }]

Icon assets

• Store the PNG alongside the metadata file.
• Reference it via logo: { "@id": "icon.png" }.
• Optional dark/light variants can use different filenames, but make sure they exist or the parser logs a warning.

Later sections walk through validation commands and checklists so your crate is accepted on the first review.