AI Platform releases¶

A monthly record of the new preview and GA features announced for DataRobot's managed AI Platform. Deprecation announcements are also included and link to deprecation guides, as appropriate.

February SaaS feature announcements¶

February 2025

This page provides announcements of newly released features in February 2025, available in DataRobot's SaaS multi-tenant AI Platform, with links to additional resources. From the release center, you can also access past announcements and Self-Managed AI Platform release notes.

February features¶

The following table lists each new feature:

Data¶

Manage NextGen data assets in the Registry¶

The Data page in Registry is a centralized hub for managing datasets in NextGen, allowing you to easily find, share, explore, and reuse data. Any dataset that you've added directly to the registry, you've linked to a Use Case, has been shared with you, or someone has added to a Use Case you are a member of, is displayed here. The Data Registry provides easy access to the data needed to address a business problem while ensuring security, compliance, and consistency.

To access the Data Registry, in NextGen, open Registry and click Data. From here, you can view, share, and delete data.

Then, click on an individual dataset to explore a dataset preview, metadata, and insights, as well as version history and related activity.

Additional improvements added to data prep in Workbench¶

This release introduces the following updates to the data preparation experience in Workbench:

• Before you’ve added data to a Use Case, you can drag-and-drop data right onto the canvas or select a different upload option offered.

  

• In the Add data modal, you can drag-and-drop data to register it in the Data Registry.

• You can now also add data using a URL.

  

Modeling¶

Time-aware data wrangling now GA¶

With time-aware wrangling, you can create recipes of operations for time-aware data and perform time series feature engineering during the data preparation phase. This method leverages the benefits of feature engineering for datasets larger than 10GB for time-aware use cases. The GA version offers support for Snowflake, Databricks, and BigQuery connections. Postgres connections and DataRobot data registry datasets are currently preview features. Improvements to the user-defined functions interface lets create new or used saved functions to significantly improve query performance.

Universal SHAP now available for time series experiments¶

With this deployment, Workbench now offers SHAP computations for time series insights—Feature Impact, Individual Prediction Explanations, and SHAP distributions per feature. For models in time series experiments, DataRobot computes a unique set of SHAP values for each combination of primary date, forecast distance, and series ID (if present). All forecast distances are considered. Use the dropdowns to control the visualizations.

Sparsity-related tasks added to Composable ML¶

In Composable ML, there are tasks that have specific input requirements around sparsity. For greater compatibility and to more easily connect to these types of downstream tasks, you can now do conversions without custom code using two new tasks: Sparse to Dense and Dense to Sparse.

Single-view model comparison now GA¶

Released as a preview feature in September 2023, the Workbench model comparison capability is now generally available for binary classification and regression, non-time aware experiments. To simplify the iterative process of solving an ML business problem, Workbench provides a model comparison tool that allows you to compare up to three models, side-by-side, from any number of experiments within a single Use Case. Instead of having to look at each experiment individually and record metrics for later comparison, you can compare models across experiments in a single view.

The comparison Leaderboard is accessible from any project in Workbench. It can be filtered to more easily locate and select models, compare models across different insights, and view and compare metadata for the selected models. See the video for a demonstration.

Detailed Blueprint views in Classic now GA¶

Blueprints that are viewed from the Leaderboard’s Blueprint tab are, by default, a read-only, summarized view, showing only those tasks used in the final model. However, the original modeling algorithm often contains many more “branches,” which DataRobot prunes when they are not applicable to the project data and feature list. Now, you can toggle to see a detailed view while in read-only mode. Previously the feature was in preview, requiring a feature flag. It is now generally available.

Predictions and MLOps¶

View model insights in the Registry¶

For DataRobot and custom models in the Registry, the Insights tab now includes Individual Prediction Explanations and SHAP Distributions: Per Feature, in addition to Feature Impact. These insights are supported for binary classification and regression problem types.

Individual Prediction ExplanationsSHAP Distributions: Per Feature

Execution environment GA improvements¶

After you create a custom model and select an environment, you can manage the environment version to ensure it is up to date. For the model and version you want update, on the Assemble tab, navigate to the Environment section. In the Environment version menu, If a newer version of the environment is available, you can click Use latest to update the custom model to use the most recent version with a successful build:

In addition, you can click View environment version info to view the environment version, version ID, environment ID, and description:

Custom environment version information is also available in the custom model’s version details.

Enable prediction warnings for a deployment¶

Enable prediction warnings for regression model deployments on the Humility > Prediction warnings tab. Prediction warnings allow you to mitigate risk and make models more robust by identifying when predictions do not match their expected result in production. This feature detects when deployments produce predictions with outlier values, summarized in a report that returns with your predictions.

If you enable prediction warnings for a deployment, any anomalous prediction values that trigger a warning are flagged in the Predictions over time bar chart. The yellow section of the bar chart represents the anomalous predictions for a point in time. To view the number of anomalous predictions for a specific time period, hover over the point on the plot corresponding to the flagged predictions in the bar chart.

Applications¶

"Talk to my data" Agent application template¶

Use the "Talk to My Data" Agent application template to ask questions about your tabular and structured data from a .csv or database using agentic workflows. This application allows you to rapidly gain insight from complex datasets via a chat interface to upload or connect to data, ask questions, and visualize answers with insights.

Decision-makers depend on data-driven insights but are often frustrated by the time and effort it takes to get them. They dislike waiting for answers to simple questions and are willing to invest significantly in solutions that eliminate this frustration. This application directly addresses this challenge by providing a plain language chat interface to your spreadsheets and databases. It transforms raw data into actionable insights through intuitive conversation. With the power of AI, teams get faster analysis helping them make informed decisions in less time.

Install non-Python dependencies for custom applications¶

When you build a custom application, you can supply a requirements.txt file in an application source to instruct DataRobot to install Python dependencies when building an app. To install non-Python dependencies, you can now provide a build-app.sh script as part of application sources. DataRobot calls the script when you build an application for the first time. The build-app.sh script can run npm install or yarn build, allowing custom applications to support dependency installation for JavaScript-based applications.