This latest release is a final installment of some major changes to the Kraken workflow. With the ability to create multiple versions of your analysis, we've added a new section in the Analysis Results section that will allow you to see all of the versions you've created and deploy a production version to use with your Predicted Datasets. We've also added a new Data Provider to the mix.
TL; DR
Final chapter of the new Kraken workflow: Deploy Analysis Versions! Also, a new Data Provider
*** NEW ***
- Analysis Versions: After creating an analysis and refining it over several iterations, click on the Deploy tab to Publish a production version of that analysis. Publishing a version allows you to create Predicted Datasets that will use the published version to create predictions that are sent back to your Data Provider. A few rules to remember about versions:
- Only one version can be published at a time
- Predicted Datasets will always use the published version of an analysis
- If no version is published, the analysis will use the most recent version to create a Predicted Dataset
- No feeding your versions after midnight, and no getting them wet
- Run Now: You now have the option to run a prediction at any time. There is a Run Now button on the Predicted Dataset card menu (three dots in the right hand corner of the card), the Settings tab of the Predicted Dataset card, or from the Deploy tab of the Analysis Results card. Clicking Run Now will immediately pull data from the attached Data Provider, run that data through your Analysis, and then push predictions back to the Data Provider
- Microsoft SQL Server: We've added MS SQL Server as a new Data Provider. Enter your credentials and Kraken will pull a list of all of the tables stored in your MS SQL Server database
*** FIXED ***
- Fixes a bug that wasn't process a <nil> value correctly from Google Big Query
- Fixes a bug that would fail getting stats if a column had 100% null values
- Other various fixes and improvements