Add Model Predictions | API Reference

You can add predictions to existing tasks in Label Studio in your Python script.

If you would prefer to set them directly in the script, see Create a Project for an example.

This demo uses the example project from Create a Project. You would need to adjust this to match your labeling configuration.

For example, sentiment maps to <Choices name="sentiment" toName="text" choice="single"> in the labeling configuration we used earlier.

For information on finding your API key and base URL, see Authenticate and Connect to the API.

LABEL_STUDIO_URL = 'YOUR_BASE_URL'
LABEL_STUDIO_API_KEY = 'YOUR_API_KEY'
PROJECT_ID=1    # replace with your project ID 
from label_studio_sdk import LabelStudio
from label_studio_sdk.label_interface.objects import PredictionValue
client = LabelStudio(base_url=LABEL_STUDIO_URL, api_key=LABEL_STUDIO_API_KEY)
project = client.projects.get(id=PROJECT_ID)
# Get the parsed labeling interface to build valid prediction payloads
li = project.get_label_interface()
# Iterate tasks and attach predictions
for task in client.tasks.list(project=project.id, include=["id"]):
    # 'label' must match the control tag name in your labeling config
    predicted_label = li.get_control('sentiment').label(['Positive'])
    prediction = PredictionValue(
        model_version="my-super-ai",
        score=0.99,  # optional
        result=[predicted_label]
    )
    client.predictions.create(task=task.id, **prediction.model_dump())

For another example, see the Jupyter notebook example of importing pre-annotated data.