Label Studio

1.23.0 verified Mon Apr 27 auth: no python

Label Studio is an open-source data labeling platform for preparing training data for machine learning models. It supports various data types including images, text, audio, video, and time series. The current stable version is 1.23.0, with active development on a nightly build track, releasing approximately monthly.

pip install label-studio

Common errors

error ModuleNotFoundError: No module named 'label_studio_sdk' ↓

cause The SDK is not installed alongside the server package.

fix

Run pip install label-studio-sdk

error ImportError: cannot import name 'Client' from 'label_studio' ↓

cause Trying to import from the wrong package.

fix

Use from label_studio_sdk import Client

Warnings

gotcha The label-studio package (server) and label-studio-sdk (client) are separate. Installing label-studio does not automatically install the SDK. Use `pip install label-studio-sdk` for programmatic access. ↓

fix Run `pip install label-studio-sdk` to use the SDK.

deprecated The `start` command with `label-studio start` will eventually be replaced by a CLI command. Check the documentation for the latest startup procedure. ↓

fix Use `label-studio start` as documented, or `python -m label_studio`.

Imports

LabelStudio
```
from label_studio_sdk import Client
```
The SDK is used for API access; the main package provides a server.

Quickstart

Quickstart to connect and create a labeling project using the SDK.

import os
from label_studio_sdk import Client

# Connect to Label Studio
ls = Client(url='http://localhost:8080', api_key=os.environ.get('LABEL_STUDIO_API_KEY', ''))

# Get or create a project
project = ls.get_project(1) if ls.get_project(1) else ls.create_project(title='My Project')
print(project)

# Import tasks
project.import_tasks([{'data': {'text': 'Hello world'}}])