YUNA becomes even more secure
The functional core of YUNA, our CORE API, has been hardened even further. From now on, all roles, with the exception of the system admins, have read-only permission. This means that all processes can still be viewed at the same time, but are protected from changes.
The CORE API controls all central processes in YUNA. It can be used to control the individual modules such as user administration, analysis jobs, the event triggers and authentications.
YUNA becomes easier
In the new version we have, among other things, simplified the script manager. Analysis scripts can now be filtered by name. Furthermore, the YUNA standard filters are no longer displayed in a resolved way, which makes the overview much slimmer.
For the „GenLinks“ function there is now a dedicated option whether these links should be opened in a new tab or the current window.
Data Scientists and users move even closer together
The stage is set for the new Messenger function. Not only can it be used to receive system messages, users can also send information directly to each other as a message within YUNA. This eliminates the need to switch to third-party applications to exchange and/or pass on information.
In the current version, the interface is already implemented and can be activated in the configuration file.
Training models on the fly
Generating training data has been a major challenge until today. With YUNA this is now done in passing. In every dashboard the result rating can be integrated directly. Thus, the user has the possibility to generate training data by rating the results on the spot of the result presentation. Your results will improve significantly due to the better data basis, if you take it up again in supervised models.
Staging and model deployment via Git branches
An automated deployment of the analysis scripts is realized in YUNA with an attached Git repository. Now it is possible to define a specific branch. Thus, different stages (development, test, productive) or the deployment of different models on different YUNA instances can be realized via the different branches.
Better planning: always keep an eye on analyses
What happened to my analysis that was supposed to start at 11:00? Why can’t I find myself in the log? We have answered these and other questions with a new convenience function. In complex contexts, analysis jobs can start at different times than originally planned, e.g. due to dependencies on other jobs.
In the new version, the system overview now shows the planned as well as the actual start of each analysis job, in order to better plan the maintenance of the YUNA instance.
YUNA meets Git: Synchronize Git Repo with YUNA
Scripts for productive analysis jobs are managed by YUNA. Whether for ETL, data preparation, ML models or visualizations: Adjustments to the scripts as well as their results – for example through input from business users or end users – should be traceable at any time through good documentation. As of YUNA 1.24, it is now possible to connect a Git repository and synchronize it with YUNA.
YUNA users can thus – depending on their authorization level – access the Git repo of the data scientists and create scripts for the respective phases of a workflow from development to test to production. Git integration in YUNA makes sophisticated data science script deployment scenarios less error-prone and significantly easier and faster to implement.
Overview of waiting analysis jobs
In YUNA, the user can schedule the execution of each analysis in analysis jobs. In doing so, YUNA automatically controls the ranking of the respective scripts in a schedule. As of YUNA 1.24, the exact start time of a queued job is displayed. By improving traceability, the update or restart of a YUNA instance, for example, can be better planned without interrupting the run of a relevant analysis job.
Job status for easier debugging
An analysis job passes through various phases such as SCHEDULED, START, FINISHED, CANCELLED in its possibly cyclically repeating life cycle. YUNA 1.24 now displays the phase a job is in at any time. This simplifies debugging and debugging a script during the development of an analysis.
Automatic script log cleanup
YUNA offers the possibility to define optional cyclic routines for the system in the system settings. These have been extended in YUNA 1.24 with an additional action that allows the administrator to clean up the generated logs of analysis scripts in a defined time window. Thus, system resources are conserved and the overview is increased.
Simplified dashboard development with YUNA-ML
In YUNA, a dashboard is defined by YUNA’s own description language YUNA-ML, which is based on XML. For publishing, the respective script is parsed, validated and the result is finally logged. The logging has now been revised and made clearer to simplify the development of dashboards.