Skip to main content

Release notes

v0.11.1 | Oct 02, 2024​

New Features:

  • Wave App is now containerized.
  • Add property IgnoreSeparatorInQuotes which prevents splitting on the separator inside double-quotes in a CSV file.
  • Support for Appstore CA-bundle.
  • Wave App adapted to work in container with Appstore.

Improvements:

  • Bump Spring Boot to v3.3.4.
  • Bump Mojo Runtime to v2.8.6.1.
  • Bump Wave SDK to v1.5.1.
  • Upgrade Python version for Client to v3.11.
  • Resolve high security vulnerabilities in all components.
  • README and Makefile commands are updated.
  • Install Material Icons locally to work in air-gapped env.

Bug Fixes:

  • Fix issue where logs cannot be obtained immediately after creating job.

v0.11.0 | Aug 13, 2024​

New Features:

  • Add DB Scorer logs support through UUIDs.
  • Introduce Autogen annotations to handle Autogen Types.
  • Add ESCORER_REMOTE_ENABLED environment variable to enable remote monitoring and deployment.
  • Add a textbox to rename models imported from the MLOps Registry.
  • Allow importing a model from MLOps Registry with and without replacement.
  • Display Model and Other Error Counts in Rest Scorer.

Improvements:

  • Bump Spring Boot to v3.3.2.
  • Bump Mojo Runtime to v2.8.5.
  • Resolve critical and high-security vulnerabilities.
  • Transition from Ubuntu to Chainguard for the base image.
  • Refactor and update Py Client to support DB Scorer changes.
  • Change label to Model ID in Registry since it is not RFC 4122 compliant.
  • Add the Drivers tab to the navigation bar.
  • Update MC Py Client Autogen Code for Jupyter Notebook.
  • Strip leading and trailing whitespaces and forward slashes.
  • Rename Wave App secret keys to use hyphens instead of underscores.
  • Improve error handling and display in REST Scorer.
  • Improve DB Scorer Error Handling.
  • Upload a model to the h2oScorePackages directory of a selected stage in SF.
  • Add shadowJar plugin to create Fat Jar of DB Scorer.
  • Add a notification bar to show connection has been removed successfully.
  • Install Monaco editor locally to work in air-gapped env.
  • Allow resuming warehouse through UI.

Bug Fixes:

  • Fix the issue where the MLOPs button is visible when MLOPs integration is not enabled.
  • Fix issue where the model with whitespaces cannot be managed in the model registry.
  • Fix the issue where the logo redirects the user to a blank page.
  • Fix issue where disabled functionalities can still be accessible via the URL.
  • Fix the issue where the H2O logo is not visible in air-gapped containers.
  • Fix issue where models from both the stage and registry are visible in SPCS.
  • Fix the issue where authentication does not work when resuming a long-suspended eScorer instance.

v0.10.3 | Aug 23, 2024​

Bug Fixes:

  • Fixed an issue where resuming an H2O eScorer instance caused authentication errors in the Wave app.

v0.10.2 | Jun 29, 2024​

Bug Fixes:

  • Fixed an issue where the latest numpy version was not compatible with the Wave app.

v0.10.1 | Apr 03, 2024​

Improvements:

  • Allow setting DB Scorer VM Options using Helm Charts.

Bug Fixes:

  • Fixed an issue where JMX Stats for LLM Models were not getting updated.

v0.10.0 | Mar 15, 2024​

New features:

  • Upgraded JDK to 17
  • Upgraded Spring Boot to v3.1.9
  • Upgraded Wave app to Wave SDK v1.0.2
  • Models can be deployed to Remote eScorer Instances, and their stats can be monitored.
  • Python Client supports scoring Pandas DataFrames.
  • Model Registry can be filtered, sorted and searched.
  • Self-Signed Certificates can now be used with MLOps client.
  • Added initial LLM support.

Improvements:

  • AutoGen generates code with syntax highlighting.
  • Snowpark Container Services queries have been updated.
  • Snowpark Container Services can be suspended and resumed within the UI.
  • Critical security vulnerabilities have been addressed.
  • eScorerAsyncClient class in Python Client has been deprecated in favor of Client.
  • Scoring a model from Model Registry auto-fills the dropdown with model name.
  • Page layouts have been improved, and navigation panel groups are collapsed by default.
  • Added refresh button to refresh list of projects in MLOps.
  • Improved documentation and docstrings in Python Client.
  • Updated documentation on Snowflake and Snowpark Container Services.

Bug Fixes:

  • Fixed an issue where aggregate toggle in model stats causes a 500 error.
  • Fixed an issue where Snowflake configuration does not update when changing the account role and database.
  • Fixed an issue where UI is empty when using incorrect credentials in Operation Monitoring.
  • Fixed an issue where external URLs caused dark mode to be triggered and users to be redirected to the home page.
  • Fixed an issue where certain actions return errors due to outdated authentication token.
  • Fixed an issue where the DB Scorer does not terminate if there is a mismatch between the number of features and selected SQL Fields.
  • Fixed an issue where Stats dialog boxes did not give clear error messages.

v0.9.0 | Jan 9, 2024​

New features:

  • Upgraded Wave app to Wave SDK v1.0.0
  • Upgraded Wave app to Python v3.10
  • Added a dark mode option to the Wave app.
  • Enhanced Snowflake monitoring charts with new axis titles and interactive tooltips.
  • Ported Snowflake operation monitoring from wave-snowflake app.
  • Added a caching layer for Snowflake chart queries.
  • Updated the statistics widgets dashboard:
    • You can now manage JDBC connections for stat widgets.
    • You can now add titles for stats widgets.
    • You can now add note text on stat widgets.
    • Encrypted JDBC connection passwords in stats widgets JSON.

Bug fixes:

  • Gracefully handled S3 client failures in DBScorer.
  • Fixed an issue that prevented access to Snowpark Container Services when switching accounts in Snowflake.
  • Fixed an issue with redirection after REST scoring where it redirected to stats instead of Model Monitor.
  • Implemented minor UI fixes and improvements in the eScorer Wave app.
  • Fixed External Function Deployment to deploy to the specified endpoint with credentials.
  • Fixed an issue that caused the Snowflake connector to fail to establish a connection in stats dashboard.
  • Fixed a typo by renaming snowflake_servce to snowflake_service.
  • Fixed a typo by renaming statitics to statistics.

Improvements:

  • Updated the eScorer Python client API authentication documentation.

v0.8.1 | Dec 14, 2023​

Bug fixes:

  • Fixed an issue where the copyright notice in the Wave app displayed an outdated year.
  • Fixed goDBScorer.sh to handle BAD DATA errors that prevented HAMC-S3 Batch Scoring for DAI models.
  • Fixed an issue with the model monitoring dashboard displaying incorrect data.
  • Fixed an issue where the model stats page failed to load when the error message contained double quotations.
  • Fixed an issue that occurred when attempting to open the deployment dialog in the Deployments Tab in Snowflake.
  • Fixed an issue related to libcrypto library incompatibility that prevented the use of Snowflake features.
  • Fixed an issue that prevented the deployment of models to External Functions via the Snowflake Deployments tab.

Improvements:

  • Upgraded packages to address security vulnerabilities in the Wave app.

v0.8.0 | Nov 29, 2023​

New features:

  • Implemented Sagemaker feature enabling SHAP column names output in RunSagemakerBatch.
  • Added support for Google Vertex in RunGoogleVertex and enhanced Google BigQuery implementation.
  • Added support for Ubuntu 22.04 (Jammy).
  • Introduced stats dashboard with new widgets: Uniques and Percentiles.

Bug fixes:

  • Fixed an issue where the MLOps registry was disabled for instances created by other users.
  • Fixed DB Scorer for more consistent error reporting.
  • Fixed an issue where model version was not being recognized for H2O models.

Improvements:

  • Helm Chart:
    • Enabled specification of an existing PV claim.
    • Added cloud discovery annotations for the service.
  • Improved error handling for cases when the eScorer service is unresponsive.
  • Enabled reading of the ESCORER_SNOWFLAKE_SFC_ENABLED environment variable from waveapp secrets specified in app.toml.
  • Enabled reading of the ESCORER_STATS_ENABLED environment variable from waveapp secrets specified in app.toml.
  • Improved documentation in the developer guide.
  • You can now pass an external lib folder to DBScorer and RestScorer JDBC connectors.
  • Addressed security vulnerabilities with the necessary security fixes.

v0.7.0 | Oct 13, 2023​

New features:

  • Added support for model deletion directly from the Wave app UI.
  • Added support for authentication using refresh tokens in the Python client.
  • Enhanced support for Snowflake users who do not have container management permissions, allowing them to manage User-Defined Functions (UDFs).
  • Added support for listing native applications in Snowflake.

Bug fixes:

  • Fixed an issue where environment variables were not being correctly replaced in the runtime properties file.
  • Fixed an issue where the DRIVERLESS_AI_LICENSE_FILE setting was not being honored.
  • Fixed an issue related to unsupported MOJOs being uploaded.
  • Fixed an issue that prevented the disabling of cloud telemetry.
  • Snowflake:
    • Fixed an issue that resulted in errors when listing Snowflake services.
    • To ensure smooth functionality, resolved issues with Snowflake monitoring charts.
    • Corrected problems with Snowflake service endpoints that caused interruptions in accessing all services.

Improvements:

  • Helm Chart:
    • Enhanced the Helm chart publishing process, now supporting ECR (Elastic Container Registry).
    • Removed deprecated Kubernetes 1.25 APIs from hpa.yaml.
    • Updated README with helm-docs.
  • Added custom support for OIDC (OpenID Connect), specifically for Snowpark Container Services.
  • Improved and updated the documentation for the Python client.
  • Implemented a feature to disable the score button until all required files are uploaded in REST and database scoring.
  • Improved the performance of Snowflake scoring implementations.
  • Enhanced error reporting and robust handling of access token management in the Wave app UI.
  • Improved error reporting on the backend when a model with the uploaded model name already exists.
  • Enhanced license check logs to aid in debugging license-related issues.
  • Improved error reporting in database scoring.
  • Addressed security vulnerabilities with necessary security fixes.

v0.6.1 | Aug 16, 2023​

Bug fixes:

  • Fixed init container check and telemetry pod labeling issues in Helm chart.
  • Addressed a bug with the "View Logs" action button on the "View Containers" page.

v0.6.0 | Aug 11, 2023​

Improvements:

  • Helm chart:

    • Added support for configuring runAsGroup and fsGroup settings within the security context.
    • Added support for namespace override.
    • Added telemetry pod labeling.

  • Enhanced the accessibility of eScorerAsyncClient in the Python client package.

  • Improved the DBScorer bash script, with DBScorer JVM thread count update.

Security fixes:

  • Upgraded the Snowflake Python connector to the latest v3.1.0.

New features:

  • Introduced Snowflake tab with Snowpark container services.
  • Introduced eScorer H2O Cloud Telemetry.

v0.5.1 | June 12, 2023​

Improvements:

  • Improved container security context in Helm chart.

Bug fixes:

  • Fixed AttributeError: 'NoneType' object has no attribute 'strip' error on wave app startup when H2O MLOps integration is not enabled.

v0.5.0 | Apr 14, 2023​

Improvements:

  • The mojo2-runtime for batch scoring is now packaged in the image.
  • The model registry now refreshes automatically after a model is uploaded.
  • You can now upload multiple database scorer libs at a time.

Bug fixes:

  • Fixed an issue that caused a timeout error with AWS.
  • Fixed a bug when uploading a CSV with a header.

v0.4.1 | Mar 22, 2023​

New

  • Added support for Ceph / Rook FS Batch Scoring.
  • Added support for multiple caCerts in Helm.

v0.4.0 | Feb 28, 2023​

H2O eScorer Backend:

  • Feature: Implemented H2O MLOps model registry integration.
  • Bug fix: Implemented several batch scoring bug fixes for AWS S3.
  • Enhancement: Added scoring termination conditions.

H2O eScorer Wave app:

  • Batch scoring: Fixed an issue with the batch scoring properties file that caused a TypeError (incorrect type).
  • Fixed a bug that resulted in an invalid file type error when scoring CSVs in real time.
  • Enhancement: Models can now have special characters in model features. (For example, :)

Helm chart:

  • keycloakSslRequired and imagePullSecrets have been externalized.

v0.3.0 | Dec 19, 2022​

New

  • H2O MLOps model registry integration: Seamless, point-and-click MLOps model deployment in H2O eScorer.
  • The H2O eScorer app now renders response status code and content on API call failures for easier debugging.
  • Added the H2O eScorer app logger. This logs useful info, most notably the backend server address, auth URL, and app version / instance ID.
  • The H2O eScorer backend now uses Helm charts for H2O AI Hybrid Cloud.

v0.2.0 | Sep 17, 2022​

New

  • Wave app: Full dataset (csv) scorer in real-time
  • Infra: H2O cli generator GUI for creating eScorer service authentication secret

v0.1.0 | Aug 03, 2022​

New Wave app features

  • Model registry
  • Deployment
  • In-app scorers
  • ModelStats dashboard

New Python client features

  • Authenticate score a model concurrently in real-time
  • View predictions, elapsed time, and the number of rows scored
Feedback