Intelliscribe

Access Intelliscribe through client.intelliscribe in the Datature Vi SDK.

Intelliscribe runs the platform's automatic annotation models against a dataset's assets, producing draft annotations that a human reviewer can accept, edit, or reject. The resource exposes a single generate() endpoint that kicks off a generation run.

Methods

generate()

Trigger Intelliscribe to generate predictions for a dataset.

result = client.intelliscribe.generate(
    dataset_id="dataset_abc123",
    body={"spec": {"taskType": "phrase-grounding"}},
)

print(result)

result = client.intelliscribe.generate(
    dataset_id="dataset_abc123",
    body={"spec": {"taskType": "vqa"}},
)

for prediction in result:
    print(prediction)

from vi.client.errors import ViOperationError, ViValidationError

try:
    result = client.intelliscribe.generate(
        dataset_id="dataset_abc123",
        body={"spec": {"taskType": "phrase-grounding"}},
    )
    print(f"Generated {len(result)} predictions")
except ViValidationError as err:
    print(f"Invalid request: {err}")
except ViOperationError as err:
    print(f"Generation failed: {err}")

Returns: list. Prediction outputs. The element shape depends on the task type (VQA, phrase grounding, or caption).

Raises: ViValidationError if dataset_id is invalid or body is empty. ViOperationError if the API returns an unexpected response shape.

list()

Response format

generate() returns a list of prediction documents. The shape of each element depends on the task type requested.

Related resources