use control.label() instead of lower level fns

src/label_studio_sdk/label_interface/control_tags.py

@@ -505,13 +505,6 @@ class SpanSelectionOffsets(SpanSelection):
 class ChoicesValue(BaseModel):
     choices: List[str]
 
-    # I don't know how Choices predictions with choice != 'multiple' was working without this...
-    @validator("choices", pre=True, always=True)
-    def coerce_to_list(cls, value: Union[str, List[str]]):
-        if isinstance(value, str):
-            return [value]
-        return value
-
 
 class ChoicesTag(ControlTag):
     """ """
@@ -547,15 +540,6 @@ def to_json_schema(self):
             "description": f"Choices for {self.to_name[0]}"
         }
 
-    def _validate_labels(self, labels):
-        if super()._validate_labels(labels):
-            return True
-
-        # HACK to continue to support single-item output in json schema
-        if not self.is_multiple_choice and isinstance(labels, str):
-            return super()._validate_labels([labels])
-
-
 
 class LabelsValue(SpanSelection):
     labels: List[str]

src/label_studio_sdk/label_interface/interface.py

@@ -840,23 +840,64 @@ def generate_sample_task(self, mode="upload", secure_mode=False):
 
         return task
 
-    def generate_sample_prediction(self):
-        """ """
+    def generate_sample_prediction(self) -> Optional[dict]:
+        """Generates a sample prediction that is valid for this label config.
+
+        Example:
+            {'model_version': 'sample model version',
+             'score': 0.0,
+             'result': [{'id': 'e7bd76e6-4e88-4eb3-b433-55e03661bf5d',
+               'from_name': 'sentiment',
+               'to_name': 'text',
+               'type': 'choices',
+               'value': {'choices': ['Neutral']}}]}
+
+        NOTE: `id` field in result is not required when importing predictions; it will be generated automatically.
+        """
         prediction = PredictionValue(
             model_version='sample model version',
             result=[
-                {
-                    'from_name': control.name,
-                    'to_name': control.to_name[0],
-                    'type': control.tag,
-                    # TODO: put special case for choices in generation instead of validation
-                    'value': {control._label_attr_name: JSF(control.to_json_schema()).generate()}
-                } for control in self.controls
+                control.label(JSF(control.to_json_schema()).generate())
+                for control in self.controls
             ]
         )
         prediction_dct = prediction.model_dump()
-        assert self.validate_prediction(prediction_dct), 'could not generate a sample prediction'
-        return prediction_dct
+        if self.validate_prediction(prediction_dct):
+            return prediction_dct
+        else:
+            logger.debug(f'Sample prediction {prediction_dct} failed validation for label config {self.config}')
+            return None
+
+    def generate_sample_annotation(self) -> Optional[dict]:
+        """Generates a sample annotation that is valid for this label config.
+
+        Example:
+            {'was_cancelled': False,
+             'ground_truth': False,
+             'lead_time': 0.0,
+             'result_count': 0,
+             'completed_by': -1,
+             'result': [{'id': 'b05da11d-3ffc-4657-8b8d-f5bc37cd59ac',
+               'from_name': 'sentiment',
+               'to_name': 'text',
+               'type': 'choices',
+               'value': {'choices': ['Negative']}}]}
+
+        NOTE: `id` field in result is not required when importing predictions; it will be generated automatically.
+        """
+        annotation = AnnotationValue(
+            completed_by=-1,  # annotator's user id
+            result=[
+                control.label(JSF(control.to_json_schema()).generate())
+                for control in self.controls
+            ]
+        )
+        annotation_dct = annotation.model_dump()
+        if self.validate_annotation(annotation_dct):
+            return annotation_dct
+        else:
+            logger.debug(f'Sample annotation {annotation_dct} failed validation for label config {self.config}')
+            return None
 
     #####
     ##### COMPATIBILITY LAYER