v0.7.0

v0.7.0

Breaking

• NewDataset's entity_defs field is now an Option for consistency

Changed

• When uploading annotated comments, empty lists of assigned / dismissed labels
  are serialized in the request. Previously empty lists were skipped which
  meant it was not possible to remove labellings (N.B. the API distinguishes
  between missing field -- labellings are unmodified -- or and empty list --
  labellings are removed).
• All *Id types now implement Hash, PartialEq, and Eq
• NewDataset and NewSource now implement Default

Added

• update source: update an existing source
• update dataset: update an existing dataset

v0.6.0

v0.6.0

Breaking

• The create bucket flag --transform-tag is now required.

Changed

• delete bulk: slight performance optimisations.
• create dataset: Accept an optional --model-family and --copy-annotations-from for the new dataset.

v0.5.3

• delete bulk: For deleting multiple comments by source id. When run with --include-annotated=false, skips annotated comments. When run with --include-annotated=true, deletes all comments.

v0.5.2

• Add support for using an HTTP proxy for all requests. The proxy configuration is saved as part of the context. Additionally, the proxy can be overridden / specified as a one off using a global command line argument --proxy https://proxy.example (e.g. similar to --endpoint).

v0.5.1

v0.5.1

v0.5.0

Breaking Changes

The -e flag used to pass in entity kinds at dataset creation has been re-purposed. One now needs to pass in a json object containing the corresponding EntityDef to be added to the new dataset. Example:

re create dataset org/example-dataset -s org/example-source --has-sentiment false -e '[{"name":"trainable_org","title":"Custom Organisation","inherits_from":["org"],"trainable":true}','{"name":"non_trainable_person","title":"Basic Person","inherits_from":["person"],"trainable":false}]'

v0.4.1

release: 0.4.1