Score models¶
Status: Current | Last reviewed: 2026-06-26
Score models are trained prediction models that workflow AI Prediction nodes can call. Open Settings → Score Models to create a model, upload training data, choose the prediction target, train the model, and copy the model ID for workflow use.
In a workflow, add an AI Prediction node, select the score model, and map workflow values into the model's feature names. The feature mapper supports adding rows one by one or pasting mappings in bulk.
Dataset CSV¶
On the model detail page, use Download template to get a small example CSV, or upload your own CSV with the same shape:
- The first row must contain unique column headers.
- Include one target column to predict and at least one feature column.
- Each remaining row should be one training example with the target populated.
- Keep values consistent inside each column, such as numbers, dates, true/false values, or labels.
After upload, the Dataset tab previews detected columns and sample values. Use the Schema tab to choose the target column, select feature columns, and pick classification or regression before training.
Test a trained model¶
The Endpoint tab includes a prediction tester after the model is trained. It shows one input field for each selected feature column and keeps the tester disabled while the model is still draft or training. After you run a prediction, the result appears as a predicted value for classification models or a regression value for regression models, with confidence and model version when available.