Deployed models
A deployed model is a trained model made available for inference and explanations.
It references the model being served and the deployment configuration that defines how it is served.
Deploy a model
Deploy a model when you have a trained model and a deployment configuration ready.
Save the returned deployed model id. You will use it to monitor the deployment, query deployed model views, rename the deployment, or delete it when it is no longer needed.
Monitor deployment status
Retrieve a deployed model to check whether it is available for inference.
A deployed model is ready for inference when its status is RUNNING.
Retrieve deployed models
Retrieve deployed models when you need to find an existing deployment.
Update a deployed model
Update a deployed model when you need to rename it.
Delete a deployed model
Delete a deployed model when it should no longer be available for inference.
Views
Deployed model views help inspect inference results, explanations, diagnostics, and evaluation outputs.
In the platform, predictions and explanations are returned as view outputs. To query a deployed model, generate a deployed model view such as the Results view and include the query rows in the view request.
- Results shows the raw explanation for a query.
- Feature attribution shows how input features contributed to a prediction.
- Module attribution shows how modules and activated rules contributed to a prediction.
- Decision actions shows decisions, actions, context, and explanations.
- Module dependency shows relationships between modules, rules, and features.
- Anomalies helps detect and understand anomalous inputs or outputs.
- Strengths and weaknesses shows error, strength, and uncertainty metrics.
- Classification evaluation evaluates classification performance.
- Regression evaluation evaluates regression performance.

