PIpioneer
dataset-analysis

Grow Dataset

Changed on

Grow an existing dataset by generating new synthetic examples.

Takes an existing classification or NER dataset and generates additional examples to reach the target size. Supports class balancing to ensure equal representation of each class.

Args: request: Growth request with dataset ID, target size, and options. auth: Authentication result with user ID and client.

Returns: GrowDatasetResponse with new dataset info and generation statistics.

Raises: HTTPException: If dataset not found, invalid type, or generation fails.

post/v1/dataset/grow

Request

Request body

dataset_idstring required

ID of the existing dataset to grow from

new_dataset_namestring required

Name for the new grown dataset

target_sizeinteger required

Total number of examples in the final dataset

class_balanceboolean

If True, balance classes equally. If False, generate equal examples per source example.

domain_descriptionstring nullable

Override domain description for generation context

temperaturenumber

Generation temperature

session_idstring nullable

Session ID for log streaming

Response

Successful Response

successboolean required

Whether the operation succeeded

datasetobject required

New dataset info (id, name, path, version)

original_sizeinteger required

Number of examples in the original dataset

new_sizeinteger required

Total number of examples in the new dataset

generated_countinteger required

Number of new examples generated

distributionobject required

Number of examples generated per class/label

token_usageinteger nullable

Estimated tokens used for generation

Changes