DatasetApi¶

class DatasetApi[source]¶

Bases: supervisely.api.module_api.UpdateableModule, supervisely.api.module_api.RemoveableModuleApi

API for working with Dataset. DatasetApi object is immutable.

Parameters

api : Api

API connection to the server.

Usage example

import os
from dotenv import load_dotenv

import supervisely as sly

# Load secrets and create API object from .env file (recommended)
# Learn more here: https://developer.supervisely.com/getting-started/basics-of-authentication
if sly.is_development():
    load_dotenv(os.path.expanduser("~/supervisely.env"))
api = sly.Api.from_env()

# Pass values into the API constructor (optional, not recommended)
# api = sly.Api(server_address="https://app.supervise.ly", token="4r47N...xaTatb")

project_id = 1951
ds = api.dataset.get_list(project_id)

Methods

convert_info_to_json	_convert_info_to_json
copy	Copies given Dataset in destination Project by ID.
copy_batch	Copy given Datasets to the destination Project by IDs.
create	Create Dataset with given name in the given Project.
exists	Checks if the dataset with the given name exists in the project.
get_free_name	Generates a free name for an entity with the given parent_id and name.
get_info_by_id	Get Datasets information by ID.
get_info_by_name	Return Dataset information by name or None if Dataset does not exist.
get_list	Returns list of dataset in the given project, or list of nested datasets in the dataset with specified parent_id.
get_list_all	List all available datasets from all available teams for the user that match the specified filtering criteria.
get_list_all_pages	Get list of all or limited quantity entities from the Supervisely server.
get_list_all_pages_generator	This generator function retrieves a list of all or a limited quantity of entities from the Supervisely server, yielding batches of entities as they are retrieved
get_list_idx_page_async	Get the list of items for a given page number.
get_list_page_generator_async	Yields list of images in dataset asynchronously page by page.
get_nested	Returns a list of all nested datasets in the specified dataset.
get_or_create	Checks if Dataset with given name already exists in the Project, if not creates Dataset with the given name.
get_tree	Returns a tree of all datasets in the project as a dictionary, where the keys are the DatasetInfo objects and the values are dictionaries containing the children of the dataset.
info_sequence	NamedTuple DatasetInfo information about Dataset.
info_tuple_name	NamedTuple name - DatasetInfo.
move	Moves given Dataset in destination Project by ID.
move_batch	Moves given Datasets to the destination Project by IDs.
move_to_dataset	Moves dataset with specified ID to the dataset with specified destination ID.
remove	Remove an entity with the specified ID from the Supervisely server.
remove_batch	Remove entities with given IDs from the Supervisely server.
remove_permanently	!!! WARNING !!! Be careful, this method deletes data from the database, recovery is not possible.
tree	Yields tuples of (path, dataset) for all datasets in the project.
update	Update Dataset information by given ID.
update_custom_data	Update custom data for Dataset by given ID.

Attributes

MAX_WAIT_ATTEMPTS	Maximum number of attempts that will be made to wait for a certain condition to be met.
WAIT_ATTEMPT_TIMEOUT_SEC	Number of seconds for intervals between attempts.

InfoType¶

alias of supervisely.api.module_api.DatasetInfo

classmethod convert_info_to_json(info)¶

_convert_info_to_json

Return type

Dict

copy(dst_project_id, id, new_name=None, change_name_if_conflict=False, with_annotations=False)[source]¶

Copies given Dataset in destination Project by ID.

Parameters

dst_project_id : int

Destination Project ID in Supervisely.

id : int

ID of copied Dataset.

new_name : str, optional

New Dataset name.

change_name_if_conflict : bool, optional

Checks if given name already exists and adds suffix to the end of the name.

with_annotations : bool, optional

If True copies Dataset with annotations, otherwise copies just items from Dataset without annotation.

Returns

Information about Dataset. See info_sequence

Return type

DatasetInfo

Usage example

import supervisely as sly

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

dst_proj_id = 1982
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 0

new_ds = api.dataset.copy(dst_proj_id, id=2540, new_name="banana", with_annotations=True)
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 1

copy_batch(dst_project_id, ids, new_names=None, change_name_if_conflict=False, with_annotations=False)[source]¶

Copy given Datasets to the destination Project by IDs.

Parameters

dst_project_id : int

Destination Project ID in Supervisely.

ids : List[int]

IDs of copied Datasets.

new_names : List[str], optional

New Datasets names.

change_name_if_conflict : bool, optional

Checks if given name already exists and adds suffix to the end of the name.

with_annotations : bool, optional

If True copies Datasets with annotations, otherwise copies just items from Datasets without annotations.

Raises

RuntimeError if can not match “ids” and “new_names” lists, len(ids) != len(new_names)

Returns

Information about Datasets. See info_sequence

Return type

List[DatasetInfo]

Usage example

import supervisely as sly

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

dst_proj_id = 1980
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 0

ds_ids = [2532, 2557]
ds_names = ["lemon_test", "kiwi_test"]

copied_datasets = api.dataset.copy_batch(dst_proj_id, ids=ds_ids, new_names=ds_names, with_annotations=True)
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 2

create(project_id, name, description='', change_name_if_conflict=False, parent_id=None)[source]¶

Create Dataset with given name in the given Project.

Parameters

project_id : int

Project ID in Supervisely where Dataset will be created.

name : str

Dataset Name.

description : str, optional

Dataset description.

change_name_if_conflict : bool, optional

Checks if given name already exists and adds suffix to the end of the name.

parent_id : Optional[int]

Parent Dataset ID. If set to None, then the Dataset will be created at the top level of the Project, otherwise the Dataset will be created in a specified Dataset.

Returns

Information about Dataset. See info_sequence

Return type

DatasetInfo

Usage example

import supervisely as sly

project_id = 116482

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

ds_info = api.dataset.get_list(project_id)
print(len(ds_info)) # 1

new_ds = api.dataset.create(project_id, 'new_ds')
new_ds_info = api.dataset.get_list(project_id)
print(len(new_ds_info)) # 2

exists(project_id, name, parent_id=None)[source]¶

Checks if the dataset with the given name exists in the project. If parent_id is not None, the search will be performed in the specified Dataset.

Parameters

project_id : int

Project ID in which the Dataset is located.

name : str

Dataset name.

parent_id : Union[int, None]

Parent Dataset ID. If the Dataset is not nested, then the value is None.

Returns

True if the dataset exists, False otherwise.

Return type

bool

get_free_name(parent_id, name)¶

Generates a free name for an entity with the given parent_id and name. Adds an increasing suffix to original name until a unique name is found.

Parameters

parent_id : int

ID of the parent entity.

name : str

Name of the entity.

Returns

Returns free name.

Return type

str

Usage example

import supervisely as sly

# You can connect to API directly
address = 'https://app.supervise.ly/'
token = 'Your Supervisely API Token'
api = sly.Api(address, token)

# Or you can use API from environment
os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()


name = "IMG_0315.jpeg"
dataset_id = 55832
free_name = api.image.get_free_name(dataset_id, name)
print(free_name) # IMG_0315_001.jpeg

get_info_by_id(id, raise_error=False)[source]¶

Get Datasets information by ID.

Parameters

id : int

Dataset ID in Supervisely.

Returns

Information about Dataset. See info_sequence

Return type

DatasetInfo

Usage example

import supervisely as sly

dataset_id = 384126

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

ds_info = api.dataset.get_info_by_id(dataset_id)

get_info_by_name(project_id, name, fields=None, parent_id=None)[source]¶

Return Dataset information by name or None if Dataset does not exist. If parent_id is not None, the search will be performed in the specified Dataset. Otherwise the search will be performed at the top level of the Project.

Parameters

project_id : int

Project ID in which the Dataset is located.

name : str

Dataset name.

fields : List[str], optional

List of fields to return. If None, then all fields are returned.

parent_id : Union[int, None]

Parent Dataset ID. If the Dataset is not nested, then the value is None.

Returns

Information about Dataset. See info_sequence

Return type

Union[DatasetInfo, None]

get_list(project_id, filters=None, recursive=False, parent_id=None)[source]¶

Returns list of dataset in the given project, or list of nested datasets in the dataset with specified parent_id. To get list of all datasets including nested, recursive parameter should be set to True. Otherwise, the method will return only datasets in the top level.

Parameters

project_id : int

Project ID in which the Datasets are located.

filters : List[dict], optional

List of params to sort output Datasets.

recursive : bool, optional

If True, returns all Datasets from the given Project including nested Datasets.

parent_id : Optional[int]

Parent Dataset ID. If set to None, the search will be performed at the top level of the Project, otherwise the search will be performed in the specified Dataset.

Returns

List of all Datasets with information for the given Project. See info_sequence

Return type

List[DatasetInfo]

Usage example

import supervisely as sly

project_id = 1951

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()
ds = api.dataset.get_list(project_id)

print(ds)
# Output: [
#     DatasetInfo(id=2532,
#                 name="lemons",
#                 description="",
#                 size="861069",
#                 project_id=1951,
#                 images_count=6,
#                 items_count=6,
#                 created_at="2021-03-02T10:04:33.973Z",
#                 updated_at="2021-03-10T09:31:50.341Z",
#                 reference_image_url="http://app.supervise.ly/z6ut6j8bnaz1vj8aebbgs4-public/images/original/...jpg"),
#                 DatasetInfo(id=2557,
#                 name="kiwi",
#                 description="",
#                 size="861069",
#                 project_id=1951,
#                 images_count=6,
#                 items_count=6,
#                 created_at="2021-03-10T09:31:33.701Z",
#                 updated_at="2021-03-10T09:31:44.196Z",
#                 reference_image_url="http://app.supervise.ly/h5un6l2bnaz1vj8a9qgms4-public/images/original/...jpg")
# ]

get_list_all(filters=None, sort=None, sort_order=None, per_page=None, page='all')[source]¶

List all available datasets from all available teams for the user that match the specified filtering criteria.

Parameters

filters : List[Dict[str, str]], optional

List of parameters for filtering the available Datasets. Every Dict must consist of keys: - ‘field’: Takes values ‘id’, ‘projectId’, ‘workspaceId’, ‘groupId’, ‘createdAt’, ‘updatedAt’ - ‘operator’: Takes values ‘=’, ‘eq’, ‘!=’, ‘not’, ‘in’, ‘!in’, ‘>’, ‘gt’, ‘>=’, ‘gte’, ‘<’, ‘lt’, ‘<=’, ‘lte’ - ‘value’: Takes on values according to the meaning of ‘field’ or null

sort : str, optional

Specifies by which parameter to sort the project list. Takes values ‘id’, ‘name’, ‘size’, ‘createdAt’, ‘updatedAt’

sort_order : str, optional

Determines which value to list from.

per_page : int, optional

Number of first items found to be returned. ‘None’ will return the first page with a default size of 20000 datasets.

page : Union[int, Literal["all"]], optional

Page number, used to retrieve the following items if the number of them found is more than per_page. The default value is ‘all’, which retrieves all available datasets. ‘None’ will return the first page with datasets, the amount of which is set in param ‘per_page’.

Returns

Search response information and ‘DatasetInfo’ of all datasets that are searched by a given criterion.

Return type

dict

Usage example

import supervisely as sly
import os

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

filter_1 = {
    "field": "updatedAt",
    "operator": "<",
    "value": "2023-12-03T14:53:00.952Z"
}
filter_2 = {
    "field": "updatedAt",
    "operator": ">",
    "value": "2023-04-03T14:53:00.952Z"
}
filters = [filter_1, filter_2]
datasets = api.dataset.get_list_all(filters)
print(datasets)
# Output:
# {
#     "total": 2,
#     "perPage": 20000,
#     "pagesCount": 1,
#     "entities": [ DatasetInfo(id = 16,
#                       name = 'ds1',
#                       description = None,
#                       size = '861069',
#                       project_id = 22,
#                       images_count = None,
#                       items_count = None,
#                       created_at = '2020-04-03T13:43:24.000Z',
#                       updated_at = '2020-04-03T14:53:00.952Z',
#                       reference_image_url = None,
#                       team_id = 2,
#                       workspace_id = 2),
#                   DatasetInfo(id = 17,
#                       name = 'ds1',
#                       description = None,
#                       size = '1177212',
#                       project_id = 23,
#                       images_count = None,
#                       items_count = None,
#                       created_at = '2020-04-03T13:43:24.000Z',
#                       updated_at = '2020-04-03T14:53:00.952Z',
#                       reference_image_url = None,
#                       team_id = 2,
#                       workspace_id = 2
#                       )
#                 ]
# }

get_list_all_pages(method, data, progress_cb=None, convert_json_info_cb=None, limit=None, return_first_response=False)¶

Get list of all or limited quantity entities from the Supervisely server.

Parameters

method : str

Request method name

data : dict

Dictionary with request body info

progress_cb : Progress, optional

Function for tracking download progress.

convert_json_info_cb : Callable, optional

Function for convert json info

limit : int, optional

Number of entity to retrieve

return_first_response : bool, optional

Specify if return first response

get_list_all_pages_generator(method, data, progress_cb=None, convert_json_info_cb=None, limit=None, return_first_response=False)¶

This generator function retrieves a list of all or a limited quantity of entities from the Supervisely server, yielding batches of entities as they are retrieved

Parameters

method : str

Request method name

data : dict

Dictionary with request body info

progress_cb : Progress, optional

Function for tracking download progress.

convert_json_info_cb : Callable, optional

Function for convert json info

limit : int, optional

Number of entity to retrieve

return_first_response : bool, optional

Specify if return first response

async get_list_idx_page_async(method, data)¶

Get the list of items for a given page number. Page number is specified in the data dictionary.

Parameters

method : str

Method to call for listing items.

data : dict

Data to pass to the API method.

Returns

List of items.

Return type

Tuple[int, List[NamedTuple]]

async get_list_page_generator_async(method, data, pages_count=None, semaphore=None)¶

Yields list of images in dataset asynchronously page by page.

Parameters

method : str

Method to call for listing items.

data : dict

Data to pass to the API method.

pages_count : int, optional

Preferred number of pages to retrieve if used with a per_page limit. Will be automatically adjusted if the pagesCount differs from the requested number.

semaphore : asyncio.Semaphore, optional

Semaphore for limiting the number of simultaneous requests.

kwargs

Additional arguments.

Returns

List of images in dataset.

Return type

AsyncGenerator[List[ImageInfo]]

Usage example

import supervisely as sly
import asyncio

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

method = 'images.list'
data = {
    'datasetId': 123456
}

loop = sly.utils.get_or_create_event_loop()
images = loop.run_until_complete(api.image.get_list_generator_async(method, data))

get_nested(project_id, dataset_id)[source]¶

Returns a list of all nested datasets in the specified dataset.

Parameters

project_id : int

Project ID in which the Dataset is located.

dataset_id : int

Dataset ID for which the nested datasets are returned.

Returns

List of nested datasets.

Return type

List[DatasetInfo]

Usage example

import supervisely as sly

api = sly.Api.from_env()

project_id = 123
dataset_id = 456

datasets = api.dataset.get_nested(project_id, dataset_id)
for dataset in datasets:
    print(dataset.name, dataset.id) # Output: ds1 123

get_or_create(project_id, name, description='', parent_id=None)[source]¶

Checks if Dataset with given name already exists in the Project, if not creates Dataset with the given name. If parent id is specified then the search will be performed in the specified Dataset, otherwise the search will be performed at the top level of the Project.

Parameters

project_id : int

Project ID in Supervisely.

name : str

Dataset name.

description : str, optional

Dataset description.

parent_id : Union[int, None]

Parent Dataset ID. If set to None, then the Dataset will be created at the top level of the Project, otherwise the Dataset will be created in a specified Dataset.

Returns

Information about Dataset. See info_sequence

Return type

DatasetInfo

Usage example

import supervisely as sly

project_id = 116482

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

ds_info = api.dataset.get_list(project_id)
print(len(ds_info)) # 1

api.dataset.get_or_create(project_id, 'ds1')
ds_info = api.dataset.get_list(project_id)
print(len(ds_info)) # 1

api.dataset.get_or_create(project_id, 'new_ds')
ds_info = api.dataset.get_list(project_id)
print(len(ds_info)) # 2

get_tree(project_id)[source]¶

Returns a tree of all datasets in the project as a dictionary, where the keys are the DatasetInfo objects and the values are dictionaries containing the children of the dataset. Recommended to use with the dataset_tree method to iterate over the tree.

Parameters

project_id : int

Project ID for which the tree is built.

Returns

Dictionary of datasets and their children.

Return type

Dict[DatasetInfo, Dict]

Usage example

import supervisely as sly

api = sly.Api.from_env()

project_id = 123

dataset_tree = api.dataset.get_tree(project_id)
print(dataset_tree)
# Output:
# {
#     DatasetInfo(id=2532, name="lemons", description="", ...: {
#         DatasetInfo(id=2557, name="kiwi", description="", ...: {}
#     }
# }

static info_sequence()[source]¶

NamedTuple DatasetInfo information about Dataset.

Example

DatasetInfo(id=452984,
            name='ds0',
            description='',
            size='3997776',
            project_id=118909,
            images_count=11,
            items_count=11,
            created_at='2021-03-03T15:54:08.802Z',
            updated_at='2021-03-16T09:31:37.063Z',
            reference_image_url='https://app.supervise.ly/h5un6l2bnaz1vj8a9qgms4-public/images/original/K/q/jf/...png'),
            team_id=1,
            workspace_id=2

static info_tuple_name()[source]¶

NamedTuple name - DatasetInfo.

move(dst_project_id, id, new_name=None, change_name_if_conflict=False, with_annotations=False)[source]¶

Moves given Dataset in destination Project by ID.

Parameters

dst_project_id : int

Destination Project ID in Supervisely.

id : int

ID of moved Dataset.

new_name : str, optional

New Dataset name.

change_name_if_conflict : bool, optional

Checks if given name already exists and adds suffix to the end of the name.

with_annotations : bool, optional

If True moves Dataset with annotations, otherwise moves just items from Dataset without annotation.

Returns

Information about Dataset. See info_sequence

Return type

DatasetInfo

Usage example

import supervisely as sly

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

dst_proj_id = 1985
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 0

new_ds = api.dataset.move(dst_proj_id, id=2550, new_name="cucumber", with_annotations=True)
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 1

move_batch(dst_project_id, ids, new_names=None, change_name_if_conflict=False, with_annotations=False)[source]¶

Moves given Datasets to the destination Project by IDs.

Parameters

dst_project_id : int

Destination Project ID in Supervisely.

ids : List[int]

IDs of moved Datasets.

new_names : List[str], optional

New Datasets names.

change_name_if_conflict : bool, optional

Checks if given name already exists and adds suffix to the end of the name.

with_annotations : bool, optional

If True moves Datasets with annotations, otherwise moves just items from Datasets without annotations.

Raises

RuntimeError if can not match “ids” and “new_names” lists, len(ids) != len(new_names)

Returns

Information about Datasets. See info_sequence

Return type

List[DatasetInfo]

Usage example

import supervisely as sly

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

dst_proj_id = 1978
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 0

ds_ids = [2545, 2560]
ds_names = ["banana_test", "mango_test"]

movied_datasets = api.dataset.move_batch(dst_proj_id, ids=ds_ids, new_names=ds_names, with_annotations=True)
ds = api.dataset.get_list(dst_proj_id)
print(len(ds)) # 2

move_to_dataset(dataset_id, destination_dataset_id)[source]¶

Moves dataset with specified ID to the dataset with specified destination ID.

Parameters

dataset_id : int

ID of the dataset to be moved.

destination_dataset_id : int

ID of the destination dataset.

Usage example

import supervisely as sly

api = sly.Api.from_env()

dataset_id = 123
destination_dataset_id = 456

api.dataset.move_to_dataset(dataset_id, destination_dataset_id)

Return type

None

remove(id)¶

Remove an entity with the specified ID from the Supervisely server.

Parameters

id : int

Entity ID in Supervisely

remove_batch(ids, progress_cb=None)¶

Remove entities with given IDs from the Supervisely server.

Parameters

ids : List[int]

IDs of entities in Supervisely.

progress_cb : Callable

Function for control remove progress.

remove_permanently(ids, batch_size=50, progress_cb=None)[source]¶

!!! WARNING !!! Be careful, this method deletes data from the database, recovery is not possible.

Delete permanently datasets with given IDs from the Supervisely server. All dataset IDs must belong to the same team. Therefore, it is necessary to sort IDs before calling this method.

Parameters

ids : Union[int, List]

IDs of datasets in Supervisely.

batch_size : int, optional

The number of entities that will be deleted by a single API call. This value must be in the range 1-50 inclusive, if you set a value out of range it will automatically adjust to the boundary values.

progress_cb : Callable, optional

Function for control delete progress.

Returns

A list of response content in JSON format for each API call.

Return type

List[dict]

tree(project_id)[source]¶

Yields tuples of (path, dataset) for all datasets in the project. Path of the dataset is a list of parents, e.g. [“ds1”, “ds2”, “ds3”]. For root datasets, the path is an empty list.

Parameters

project_id : int

Project ID in which the Dataset is located.

Returns

Generator of tuples of (path, dataset).

Return type

Generator[Tuple[List[str], DatasetInfo], None, None]

Usage example

import supervisely as sly

api = sly.Api.from_env()

project_id = 123

for parents, dataset in api.dataset.tree(project_id):
    parents: List[str]
    dataset: sly.DatasetInfo
    print(parents, dataset.name)


# Output:
# [] ds1
# ["ds1"] ds2
# ["ds1", "ds2"] ds3

update(id, name=None, description=None, custom_data=None)[source]¶

Update Dataset information by given ID.

Parameters

id : int

Dataset ID in Supervisely.

name : str, optional

New Dataset name.

description : str, optional

New Dataset description.

custom_data : Dict[Any, Any], optional

New custom data.

Returns

Information about Dataset. See info_sequence

Return type

DatasetInfo

Usage example

import supervisely as sly

dataset_id = 384126

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

new_ds = api.dataset.update(dataset_id, name='new_ds', description='new description')

update_custom_data(id, custom_data)[source]¶

Update custom data for Dataset by given ID. Custom data is a dictionary that can store any additional information about the Dataset.

Parameters

id : int

Dataset ID in Supervisely.

custom_data : Dict[Any, Any]

New custom data.

Returns

Information about Dataset. See info_sequence

Return type

DatasetInfo

Usage example

import supervisely as sly

dataset_id = 384126

os.environ['SERVER_ADDRESS'] = 'https://app.supervisely.com'
os.environ['API_TOKEN'] = 'Your Supervisely API Token'
api = sly.Api.from_env()

new_ds = api.dataset.update_custom_data(dataset_id, custom_data={'key': 'value'})