Release Notes | xpresso.ai

Introduced public and protected modes for deployed service's URL.
The ability to export a solution from one xpresso instance and
import it to another instance.
- Known Limitation: DataSource import/export is not supported at the moment.
Introduced support for running only some specific components in a
pipeline.
- Known Limitation: Only for Kubeflow runs.
Introduced support for restarting a failed pipeline run from the
component where it initially failed.
Introduced Visual Studio Code as an online IDE.
- Known Limitation: The current approach isn't fully secure.
Improved support for multiple deployment clusters in one instance.
- Known Limitation: Cluster allotment for old users is a manual step
  that will be done by the xpresso.ai team while upgrading
  versions.
For new users, clusters will be allotted at the time of user
creation.
Improvements in ML Components (improved metrics and graph
generation, improved status reporting, etc.)
- Known Limitation: Only SKLearn is done as of now. Keras, LightGBM
  and XGBoost to follow in the next release.
Improved the data flow between components.

Pipeline Wizard introduced simplifying the end-to-end pipeline run process.
Introduced support for running PySpark Model Deployment and Inferencing on Airflow components.
System Check APIs introduced for all microservices (except Jupyterhub)to get a quick health report of the microservice and the external services it depends on.
Enhanced the xpresso Logger to enable passing custom labels to the logs for a better log search experience in Kibana

Made improvements to the Model Monitoring user experience.
Improved the build version display on the UI(now branch name, commit ID and descriptions Are also displayed).
Pipeline Wizard backend deployed, UI is WIP.
Improved the organization of sample solutions on the UI.

Introduced Inference.
Pipeline Service component for easy conversion of a pipeline to an Inference Service.
Improved the data flow between pipeline components.
Revamped the status checking mechanism for jobs and databases.
Added support for Azure Files in DataOps.
Introduced provision to provide the branch_name as a parameter while pushing the experiment output to the Data Repository.
Increased the number of search results from 1000 to 2000 for the list_branch API.

Improved data flow between components in a Kubeflow pipeline.
Data Source library calls.
Support for heterogeneous component types/flavors in an Airflow pipeline.
Improved build speed because of pre-build component base images.
Improved build status page because of replacing Jenkins API calls with database calls.
Revamped the status checking mechanism for services and inference services.

Introduced support for Airflow Spark pipeline.
Logger Improvements
- Users can now search for their logs on the Kibana dashboard.
- PySpark logs are also aggregated in the same way for consistency.
xpresso_ai component library can now be installed via pip using our local pypi server.
Jupyter notebooks can now be used in user workspaces.

Provision to select node labels during deployment.
Old service versions will now be automatically scaled down when a new one is deployed.

Data Sources UI and library.
Provision to allow all email domains while registering new users on xpresso.ai
Data Sources:
- Multiple files can delete & download.
- Copy/Move now supported.
- Folder creation in all data sources.

See xpresso.ai’s Control Center In Action