next

Edit

Version 10.2.2

Release date

August 13, 2021


The release introduces the following major improvements to IA Cloud Enterprise:

  • Open Development Framework 2 (ODF 2) to facilitate Use Case development.
  • Web Content Accessibility Guidelines (WCAG) compliance for WorkSpace to enhance the WorkFusion product usability, including for people with disabilities.
  • Sensitivity analysis to improve the quality of developed ML models and feature importance explanations to provide visibility into their implementation.
  • Support of multiple Python environments to enable greater flexibility in building and running Python-based ML models.

Within the release scope, the following features were ported from version 10.1.6.2:

  • a new SKIP_datastores strategy to avoid data losses during Asset Bundle import
  • a new user interface in Control Tower to simplify management of AutoML models
  • encryption of answers to Manual Tasks to enhance private data security within the platform

Other essential updates affected WorkSpace, making the app workflow more flexible, and the infrastructure and installer, enhancing the product security.

WCAG Compliance for WorkSpace

WorkSpace was updated to comply with the Web Content Accessibility Guidelines (WCAG)—an internationally recognized industry standard regulating the representation of the information on web pages or in apps. The update is immediately available—no additional configuration is needed on the client's side.

The improvement enhances the usability of the WorkFusion product for Operations specialists, including people with disabilities. Customers are enabled to meet the requirements of employment policies and upgrade the operational comfort for their employees, thus improving overall work productivity.

ODF 2

ODF 2 is the next level of a well-thought framework to support implementation teams in building complex, large but testable, and maintainable solutions. The main features include:

  • The Use Case-based approach representing an ODF project as a Use Case that encapsulates all resources required for its work.
  • The Java-based approach for standard Bot Tasks enabling to code the Bot Task logic in pure Java.
  • The ODF Bundle Versions Maven plugin developed to simplify the versioning of Use Case Asset Bundle.
  • A library of out-of-the-box extensible templates for the most common Bot Task scenarios to be reused across different projects.
  • The built-in ORMLite framework.
  • Data Model defined in Liquibase migrations allowing to create and modify Data Store tables with different column types, index constraints, and relations between the tables.
  • The out-of-the-box integration of a standard Data Model with Datamart 2.0.

For more details, refer to the ODF 2 documentation.

Explainability and sensitivity analysis for ML model

New AutoML capabilities enable developers to configure and run sensitivity analysis and feature importance explanations for ML models in development, including via Control Tower UI. For out-of-the-box models, sensitivity analysis is already pre-configured, and you can create custom configurations. Both cluster and local execution options are available.

The new features facilitate the adoption of Machine Learning models and improve their quality. Model explanations help you make models more reliable and trustable by providing better visibility into their delivery process. Sensitivity analysis allows you to identify uncertainty sources and use this knowledge to model predictions more reliably.

For details, refer to the Sensitivity Analysis and Model Explanations documentation.

Support of multiple Python environments for ML models

Earlier, IA Cloud Enterprise had no out-of-the-box mechanism enabling different Python environments by different ML Workers. Since 10.2.2, IA Cloud supports multiple Python environments:

You can now run ML models based on different Python versions (Python 2.7 or Python 3.7) in the same environment. This facilitates implementing and deploying Python-based customizations for the AutoML capability. For details, refer to the configuration guide.

Improvements

Analytics

  • Changed the size for RPA Bot Unit columns.
  • Out-of-the-box dashboards for Analytics are no longer published during the installation by default. When needed, you can install them manually. For the guide, see Enable Analytics dashboard.

AutoML

  • A basic user interface (UI) was ported from 10.1.6.2 to simplify management of AutoML artifacts, trained models, and model training. For details, refer to the AutoML UI documentation.

Bot Execution Platform

  • Introduced a mechanism for handling stuck Bot Tasks. For more details, refer to Configure Bot Task alerts and forced stop.
  • Introduced timeouts for task execution.
  • Decreased the bep.tds.broker.prefetch-count parameter’s default value from 50 to 5. The parameter sets the number of messages to be loaded into the TDS memory for each task queue TDS is listening to. The change improves stability and lowers RabbitMQ and Task Dispatcher Service memory consumption.
  • Increased the spring.rabbitmq.max-channels parameter’s default value from 2047 to 4096. The limitation is introduced to protect RabbitMQ and TDS from potential memory pressure.
  • Enabled the Lazy Queues generation by default in Control Tower. These queues move their contents to disk as early as possible and load them in RAM only when requested by consumers.
  • Introduced cleanup of task queues that are no longer required in active Business Processes.
  • Introduced the mechanism for handling errors from worker types that WMS does not manage. If a worker group fails for some reason, the error mode is activated, and an error response is sent for all tasks intended to this worker group. After fixing the problem with workers connected to the worker queue, the mode is disabled automatically, and the task broker returns to regular operation.

Infrastructure and installer

  • Changed Python version from 2.7 to v.3.5 to increase the Product security.
  • Added support of Red Hat Enterprise Linux (RHEL) 8.3 and CentOS 7.9.
  • Updated the Kerberos krb5-libs and krb5-workstation packages to v.1.18.2 to fix various vulnerability issues.
  • Provided the possibility to upgrade the Analytics server to the latest Tableau version 2020.4 for security and compliance reasons. For more information, see the Upgrade guide.
  • Fixed multiple vulnerability issues by changing the following components:
    • Apache Subversion updated to 1.14.0
    • Erlang updated to 22.3.3
    • RPCBind updated to 0.2.0-7.fc14
    • Rsync updated to 3.2.3
    • Keycloak updated to 11.0.3
    • Apache Xalan updated to v2.7.2

RPA

  • Configured Manipulation Framework timeouts for RPA Bots to avoid getting stuck in native calls. For more details, refer to RPA Manipulation Framework timeout configuration.
  • Configured a reconnection action timeout for the command-line start of the Unit Bot Agent to address the RPA-to-BEP interaction issues due to connectivity disruptions between Bot Agents and RabbitMQ. For more details, refer to Configure timeout to reconnect.
  • Modified start-worker.bat so that it contains a debug argument after IA Cloud Developer installation.

WorkSpace

The updates introduced into the WorkFusion product allow customers to:

  • Configure a custom assignment priority, making it possible to range incoming Manual Tasks in WorkSpace by the significance or other criteria. For detailed instructions, refer to the Configure custom assignment priority guide.
  • Specify assignees for individual assignments. For detailed instructions, refer to the Configure assignee for assignment guide.

The new features enable customers to flexibly arrange the workflow both on the company-wide level and on the level of individual employees based on business requirements.

In addition, the encryption of answers to Manual Tasks was ported from 10.1.6.2 to enhance private data security within the platform. For detailed configuration instructions, refer to the Encrypt Manual Task answers article.

Bug fixes

AutoML

  • Changed the unit to denote the memory usage limit for subprocesses under AutoML Workers from kilobytes to megabytes in Worker YAML files.
  • Included quick-start ML Archetypes to the installation package.

Bot Execution Platform

  • Fixed OOM exceptions in ZooKeeper caused by expired nodes in Vault.

Control Tower

  • Fixed the issue with the Business Process not being completed because of too many transition tasks on a final step.
  • Fixed the task queue issue. Now the queue is re-created after the bot source changes during Business Process pause and resume.
  • Fixed the issue with worker initialization depending on the number and size of saved workers’ logs. The worker startup is now stabilized.
  • Fixed the issue with Control Tower sending unhelpful Business Process logs to Kibana.
  • Fixed the issue with training Business Process failure when the WFML Job Data Data Store was not updated, blocking the consecutive training. The fix adds more transparency and improves troubleshooting.
  • Fixed the issue with Control Tower submitting tasks twice when TDS is not working, and a Business Process is stopped and subsequently resumed.
  • Fixed vulnerability issues related to the Bounty Castle 1.5x library.
  • Fixed the issue with Hazelcast throwing the "Partition does not have an owner. partitionId: 74" exception.
  • Fixed the issue with a Business Process stuck because of TDS losing a connection to channels.
  • Fixed the issue with a grid not displayed in the preview and external preview while creating a Manual Task.
  • Fixed the file manager appearing blank while creating a Manual Task.
  • Updated old CT dependency.
  • Fixed the Hyper Models list not displayed.
  • Fixed the data table not displayed when a secured file in a Manual Task gets opened.
  • Improved Business Process execution performance by stopping TDS from trying to consume from deleted queues.
  • Fixed the RPA worker losing connection with worker pool queue and becoming useless.
  • Fixed the issue with a task duplicated during submission to RabbitMQ task queues.

IA Cloud Developer

  • Fixed the "Login failed for user sa" error during the IA Cloud Developer installation.
  • Fixed the issue with Keycloak not starting in a two-minute interval on a slow machine.

Infrastructure and installer

  • Fixed issues with logs from the RPA machines not appearing in Kibana.
  • Fixed the issue of prechecks passing when an incorrect certificate chain was provided.
  • Fixed multiple vulnerability issues to prevent clickjacking or UI redress attacks.
  • Added Mesos/Marathon ports to prechecks to stabilize the installation.
  • Added the Keycloak log file to rotation to avoid excessive disk space consumption.
  • Reduced the number of false-positive error log messages coming from ZooKeeper and not bearing any helpful information.
  • Fixed the issue with the RPA server becoming unresponsive after the installation.
  • Added encryption to secure network communication with RabbitMQ server and prevent leakage of sensitive customer data.
  • Fixed worker-management-service getting refreshed constantly.
  • Fixed the security issue by creating a list of undesirable user agents for Nginx to return 403 error if any of them are present in the user-agent string.
  • Configured Nginx buffer overflow protection, minimizing the buffers' sizes.
  • Fixed the version conflict of Python packages.

ODF

  • Fixed the issue with Bot Configurations not being updated in a Business Process after consecutive imports.
  • Fixed the caching issue in Business Processes email settings.

RPA

  • Fixed RPA ChromeDriver getting stuck during article calls.
  • Fixed the issue with a health-check Business Process failing on the RPA step.
  • Fixed the issue with Bot Master not signing in automatically.
  • Fixed the RPA Bot autologin issue by enabling the RPA service to automatically close a security legal banner if it is configured on the RPA server and log in into the Bot Master session.
  • Fixed the issue with the RPA installation failing in the single-point mode and the RPA server becoming unreachable after executing RPAInstaller.exe.

WorkSpace

  • Fixed the document loading failure in Manual Tasks when the browser shifts to locales unsupported by WorkSpace.
  • Made the TOD field optional and prevented it from stopping a Manual Task.
  • Fixed TOD image rendering related to CT language instead of a browser locale.
  • Fixed the issue with a Manual Task not appearing in the subsequent creation of a Business Process.
  • Fixed the browser issue that prevented a user from taking an assignment.
  • Fixed the issue with Сontrol Тower not working after the upgrade from 10.1.6.1.
  • Fixed default answer labels in TOD shifting on several symbols.
  • Fixed the issue with user login activity and time not tracked in the WorkSpace database.
  • Fixed the issue with a Manual Task disappearing when the task has more than 20 business attributes.
  • Fixed the issue with the same page title for each web page of the WorkSpace application.
  • Implemented the ability to zoom documents in TOD up to 500% in the Chrome browser.
  • Changed a timestamp from the US time to local time.
  • Fixed the issue with tagged text not highlighted correctly by TOD tags selected by AutoML.

Deprecations

  • Excluded Control Tower deleted resources from the ML_STAT procedure loading.
  • Removed vulnerable JavaScript libraries.
← Version 10.2.3Version 10.2.1.3 →