Workspace and Dataset Management#

This guide explains how to set up and manage the workspaces in Argilla via the Python client.

Note

The Workspace class for workspace management has been included as of the Argilla 1.11.0 release and is not available in previous versions. But you will be able to use it with older Argilla instances, from 1.6.0 onwards, the only difference will be that the main role is now owner instead of admin.

Workspace Model#

A workspace is a “space” inside your Argilla instance where authorized users can collaborate. It is accessible through the UI and the Python client.

If you’re an owner, you can assign users to workspaces, either when you create a new user, or using the method add_user from the Workspace class.

An owner has full access to the workspace, and can assign other users to it while the admin role can only access the workspace but cannot assign other users to it, and the annotator role can only access their assigned datasets in the workspace they belong to and annotate it via the UI.

An Argilla workspace is composed of the following attributes:

Attribute	Type	Description
`id`	`UUID`	The unique identifier of the workspace.
`name`	`str`	The name of the workspace.
`inserted_at`	`datetime`	The date and time when the workspace was created.
`updated_at`	`datetime`	The date and time when the workspace was last updated.

Python client#

The Workspace class in the Python client gives developers with owner role the ability to create and manage workspaces in Argilla, and the users that belong to them. Check the Workspace - Python Reference to see the attributes, arguments, and methods of the Workspace class.

The Workspace class in Argilla is composed of the following attributes:

Attribute	Type	Description
`id`	`UUID`	The unique identifier of the workspace.
`name`	`str`	The name of the workspace.
`users`	`List[User]`	The list of users that belong to the workspace.
`inserted_at`	`datetime`	The date and time when the workspace was created.
`updated_at`	`datetime`	The date and time when the workspace was last updated.

How to work with Workspaces#

Create a new `Workspace`#

CLI

You can create a new workspace in Argilla using the create command in the workspaces group.

argilla workspaces create my-new-workspace

Python client

Creating a workspace in Argilla is now as easy as calling the create method from the Workspace class. It will return a Workspace instance.

import argilla as rg

rg.init(api_url="<ARGILLA_API_URL>", api_key="<ARGILLA_API_KEY>")

workspace = rg.Workspace.create("new-workspace")

List `Workspaces`#

The users with owner role can list all the existing workspaces in Argilla, while the users with admin role or annotator role can only list the workspaces they belong to.

CLI

You can list the workspaces in Argilla using the list command in the workspaces group.

argilla workspaces list

Python client

You can also list the workspaces in Argilla using the list method. It will return a list of Workspace instances.

import argilla as rg

rg.init(api_url="<ARGILLA_API_URL>", api_key="<ARGILLA_API_KEY>")

workspaces = rg.Workspace.list()
for workspace in workspaces:
   ...

Get a `Workspace` by name#

Python client#

You can get a workspace by its name using the from_name method. It must exist in advance in Argilla, otherwise, an exception will be raised.

import argilla as rg

rg.init(api_url="<ARGILLA_API_URL>", api_key="<ARGILLA_API_KEY>")

workspace = rg.Workspace.from_name("new-workspace")

Get a `Workspace` by id#

Python client#

Additionally, if you know the id of the workspace, you can get it directly using the from_id method. It must exist in advance in Argilla, otherwise, an exception will be raised.

Note

The id of a workspace is a UUID, and it is generated automatically when you create a new workspace.

import argilla as rg

rg.init(api_url="<ARGILLA_API_URL>", api_key="<ARGILLA_API_KEY>")

workspace = rg.Workspace.from_id("00000000-0000-0000-0000-000000000000")

Add, list, or delete users from a `Workspace`#

CLI

You can add or delete users from a workspace using the add-user and delete-user commands in the workspaces group.

argilla workspaces --name my-workspace add-user bob
argilla workspaces --name my-workspace delete-user bob

Also, you can list the users of a workspace using the list command in the users group with the --workspace option.

argilla users list --workspace my-workspace

Python client

Once you instantiate a Workspace instance from a workspace in Argilla, you can add, list, or delete users from it. But note that just the owner has sufficient permissions to perform those operations.

import argilla as rg

rg.init(api_url="<ARGILLA_API_URL>", api_key="<ARGILLA_API_KEY>")

workspace = rg.Workspace.from_name("new-workspace")

users = workspace.users
for user in users:
   ...
workspace.add_user(user.id)
workspace.delete_user(user.id)

Delete a `Workspace`#

Python client#

You can also delete a workspace using the Python client.

Note

To delete a workspace, no dataset can be linked to it. If the workspace contains any dataset, deletion will fail.

Note

You can refer to the delete datasets section below to clear a workspace before deleting it.

import argilla as rg

rg.init(api_url="<ARGILLA_API_URL>", api_key="<ARGILLA_API_KEY>")

workspace = rg.Workspace.from_name("new-workspace")

workspace.delete()

Dataset Model#

A dataset is a container of the “records” of your Argilla instance. It offers all the requirements for storing and managing the data. You can find more information about the concepts and structures of datasets here.

On the Argilla UI, you can see all the datasets you have created. A dataset is created within a workspace and it is only reachable via this specific workspace. Depending on the project, as you give access to owner, admin or annotator, you also specify which roles can reach each dataset.

The attributes of a dataset are as follows:

Attribute	Type	Description
`id`	`UUID`	The unique identifier of the dataset.
`name`	`str`	The name of the dataset.
`url`	`str`	The unique URL of the dataset.
`fields`	`list`	The TextFields that the dataset contains.
`questions`	`list`	The questions that the dataset contains.
`guidelines`	`str`	The guidelines that the dataset has.

How to work with Datasets#

You can refer to the CLI page for guidance on how to work with datasets on CLI.

Note

To work with the datasets on Python, you need to log in to Argilla with rg.init().

List `Datasets`#

Python client#

You can list the datasets within a specific workspace with the list() method as seen below. To specify the workspace, you can use the workspace argument. Otherwise, it will list all the datasets in all workspaces.

import argilla as rg

rg.init(api_url="<ARGILLA_API_URL>", api_key="<ARGILLA_API_KEY>")

dataset_list = rg.FeedbackDataset.list(workspace="admin")

for dataset in dataset_list:
   print(dataset.name)

As the list() method creates a list of RemoteFeedbackDataset objects, you can directly work each item of the list.

Delete `Datasets`#

Python client#

You can delete any dataset by pulling it from the server by from_argilla() and calling the delete() method.

rg.FeedbackDataset.from_argilla("my_dataset", workspace="admin").delete()

Workspace and Dataset Management#

Workspace Model#

Python client#

How to work with Workspaces#

Create a new Workspace#

List Workspaces#

Get a Workspace by name#

Python client#

Get a Workspace by id#

Python client#

Add, list, or delete users from a Workspace#

Delete a Workspace#

Python client#

Dataset Model#

How to work with Datasets#

List Datasets#

Python client#

Delete Datasets#

Python client#

Create a new `Workspace`#

List `Workspaces`#

Get a `Workspace` by name#

Get a `Workspace` by id#

Add, list, or delete users from a `Workspace`#

Delete a `Workspace`#

List `Datasets`#

Delete `Datasets`#