dataset describe
Show a dataset version's schema, direct dependencies and dependents, saved script and producing job. Reads metadata only.
Details describe one version: the latest complete version, or an exact version selected with name@version. Versions lists all available complete versions.
Text output
Dependencies lists direct inputs; Dependents lists datasets built from the selected version. Storage is shown as URI@listing-version. Dates use your local time zone.
JSON output
JSON includes canonical names, stable dataset_uuid and version_uuid identifiers, and UTC timestamps. The dataset UUID stays the same across versions and renames within a catalog. Unavailable sections are null.
Full JSON example, including version, dependency and job metadata.
Python
dc.describe() returns a DatasetDescription with DatasetInfo versions, a typed schema and the selected version's Job.
Datasets in the local namespace use the local catalog. Other namespaces use a pulled copy first, then Studio if the dataset or version hasn't been pulled. A copy lists only pulled versions. --studio (studio=True) skips the copy. Studio uses the team set by datachain auth team or DATACHAIN_STUDIO_TEAM. Inside Studio jobs, reads always use Studio. Source identifies the catalog used.
Studio's public API provides no dependencies or dependents, and job params is empty. Inside Studio jobs, dependencies are available; dependents are not. Pulling does not copy lineage. Unavailable sections show unavailable (null in JSON), missing inputs show unavailable dependency, and jobs without details show Job: <id> (unavailable).