New / updated manipulators
Next Test Recommender Evaluation: Test plan added as a new output of the step
The Next Test Recommender Evaluation step returns an additional dataset which contains the test plan/strategy of each recommender evaluated. Each repetition of each recommender will create a new column in the dataset specifying the order in which the different tests are selected:- 0 for the initial dataset
- 1, 2, 3, … for the first, second, third etc batches
- negative values for the data selected for the evaluation.
- Order by a specific column to view corresponding test plan
- Which test is often selected in the first few batches (more important). Which test is always selected towards the end (less important)
Anomaly Detection Finetuner: Additional information in description
Anomaly Detection Finetuner now displays the quantile and steepness values in its description when it’s run. This enables the users to know what parameters were selected to fine tune the model without having to edit the step.New features (non-manipulators)
Apps: Automate Notebook Workflows with Scheduled Runs
The new Apps feature enables users to create automated, scheduled runs of notebooks, helping to build repeatable workflows. Apps are ideal for scenarios where input data changes regularly — they support dynamic inputs by allowing the Tabular Loader to load files based on a defined file pattern. With this release, users can:- Set up schedules for notebooks to run automatically.
- Dynamically vary the input data using pattern-based file loading.
- Use Apps to streamline recurring data processing and inference workflows.
Apps currently support notebooks containing the following types of steps:
- Tabular/MF4 Loader
- Visualisation steps
- Transform steps
- Inference steps
Notebooks that contain model training steps are not supported within Apps workflows. While such notebooks can still be used to create Apps and scheduled, these steps will not execute and may cause workflow failure. This limitation is intentional for this version, and future releases may introduce support for training steps. Notebooks containing visualisations can be converted to an App, however the visualisations can not be viewed in the schedule run history. For more information on setting up Apps, see the relevant documentation.
Increased Plot Point Limit to 30,000
We’ve increased the point plot limit from 10,000 to 30,000 points across all plots previously capped. This enhancement enables more detailed visualizations, but please note it may lead to browser memory limits being reached sooner. Ensure that plot previews still render correctly where applicable.Removed CustomSQL Form Validation
In the v1.9 release, we introduced SQL query syntax validation for the CustomSQL form. However, this validation proved ineffective and often obstructed user workflows. As a result, it has now been removed to ensure smoother and more flexible usage.Updated Support Widget: Freshdesk Replaced with Service Desk
We’ve replaced the Freshdesk widget with a new Service Desk widget. While the interface looks similar, there are a few key differences in how users interact with it:- Searching Documentation: When clicking Support, users will now be prompted to type a question to search through Jira-based documentation.
- Reporting Bugs:
- Option 1: Type a question in the support widget, then navigate to Report a Bug.
- Option 2: Click your avatar in the top-right corner of the browser, select Knowledge Base, and from there either report a bug or access user documentation.
Bug Fixes
- Fixed CustomSQL step query length limits, users are no longer limited to the length of query that they use in CustomSQL steps.
Known issues in this version
Folder Deletion Regression Introduced in v1.10In versions prior to v1.10, users could delete folders even if they contained notebooks. Following the introduction of Apps, a regression was introduced that now requires folders to be empty before they can be deleted. As a result, users must first delete all notebooks within a folder before the folder itself can be removed. We are actively reviewing this behavior and plan to introduce enhancements in upcoming releases to make folder deletion more intuitive and safer across the platform.