Jump to: navigation, search

AI Core Services is part of 9.x starting in
Up to release, AI Core Services was known as Journey Optimization Platform.

AI Core Services Release Notes

Release Date Release Type Restrictions AIX Linux Solaris Windows
03/26/19 General Under Shipping Control X

Helpful Links

What's New

This release contains the following new features and enhancements:

  • This release includes an updated and improved version of the Predictive Routing API Reference. In particular, there are now cURL request examples for each endpoint.
    • This Reference requires a password for access. Please contact your Genesys representative if you need to view this document. (PRR-4144)
  • AI Core Services now requires Docker version 18.09.2, which addresses important security issues.
    • The Docker release notes for release 18.09.2 recommend the following actions:
      • Update runc to address a critical vulnerability that allows specially crafted containers to gain administrative privileges on the host. CVE-2019-5736
      • Ubuntu 14.04 customers using a 3.13 kernel will need to upgrade to a supported Ubuntu 4.x kernel.
    • The following Known Issue is noted in the Docker release notes for release 18.09.2:
      • There are important changes to the upgrade process that, if not correctly followed, can have impact on the availability of applications running on the Swarm during upgrades. These constraints impact any upgrades coming from any version before 18.09 to version 18.09 or greater.
    • For additional information, refer to the Docker blog post and the Docker release notes.
    • See the Upgrade Notes (below) for the procedure to use to stop and restart AICS when upgrading Docker. (PRR-4419)
  • You can now convert a regular account into an LDAP account. A toggle with label LDAP has been added to the Settings > Account update window. To enable LDAP authentication, enter your LDAP credentials and save changes. After that, you must also convert the user accounts for those who should use LDAP authentication. User configuration is done in the Settings > User Management window. (PRR-4385)
  • This release includes the following improvements to the user interfaces in the GPR web application:
    • A new navigation panel provides a tree view of all Datasets, Predictors, and Models configured for the current Tenant. This tree-view pane is available from the Settings > Datasets and Settings > Predictors windows. Each item in the tree view links to the specified object, enabling easy access to the entire hierarchy of Datasets, Predictors, and Models.
      NOTE: Composite Predictors, which can be built on data from multiple Datasets, are not shown in this tree-view pane. (PRR-4351)
    • For simplified navigation, breadcrumb links now appear at the top of windows in the GPR web application if you have drilled-down past a top-level window. (PRR-4350)

Resolved Issues

This release contains the following resolved issues:

You can now successfully upgrade an Account. Previously, this action generated an error message caused by an obsolete reference to a MongoDB collection that had previously been removed from AICS. (PRR-4445)

When you are preparing a Feature Analysis report for a high-cardinality target metric, you are no longer required to configure the Target Metrics <metric_name> Range field. Previously, it was mandatory to separate high-cardinality metric values into ranges. (PRR-4377)

Automatically-generated emails coming from the GPD web application or API now contain the correct link to update your password and have been rewritten to be more user-friendly (PRR-4264)

This release provides improved memory handling when you generate a Predictor. Previously, if your Predictor required extensive memory to generate the data it required for training, it might run out of memory and never finish the generation job. (PRR-4237)

If you start performing a search on a Predictor Details window, then switch to a different Predictor without clearing the Search text box, the new Predictor now displays correctly. Previously, the search term was applied to the new Predictor, which resulted in an empty data table. To display the data, you had to clear the search manually. (PRR-4173)

When you are creating a Model using the API, GPR now validates that the Customer features you specify are actually present in the specified Customer Profile schema and that you have specified at least one Customer feature. Previously, an incorrect Customer feature (one spelled incorrectly, for example), or no Customer features, resulted in an empty Model with no error message explaining the issue. (PRR-4134; PRR-2385)

When running the start.sh and restart.sh scripts to deploy an updated version of AI Core Services, you might now receive a correct message instructing you to specify the public IP address for AI Core Services in the S3_ENDPOINT environment variable, found in the tango.env file. Previously, the error message generated in this situation was incorrect, instructing you to run the install.sh script. (PRR-4124)

If you generate a Lift Estimation report via API and any column name contains spaces or non-ASCII characters, the report shown in the GPR web application now correctly displays the Group By parameter. Previously in this scenario, it displayed the hash value instead of the correct column name for those features that include spaces or non-ASCII characters in the feature name. (PRR-4102)

You can now purge data from GPR even if it contains spaces or non-ASCII characters. (PRR-4026)

If you send a request to the GPR API that contains malformed JSON (with an extra quote, for example) in the request body or non-UTF-8 encoding in the request headers, you now receive an understandable response message in return. (PRR-3997)

The Agent Profile schema now displays all rows correctly after appending data, even if the data you initially uploaded, with which GPR created the schema, included null values in some fields. (PRR-3643)

This release improves performance for displaying the AUC chart for a Model. Users with existing Models must run the upgrade_44a_predictor_auc_chart.py script, as explained in the Upgrade Notes (below) to have those Models display correctly. (PRR-3119)

You can now override a feature from a Dataset with a feature having the same name from the Agent Profile or Customer Profile.

In addition, scoring now functions correctly if you accidentally upload a Dataset and an Agent Profile or a Customer Profile that have features with the same name but with different data types. For example, if the Dataset has a feature with the integer data type and it is used in a Predictor, while an Agent Profile or a Customer Profile has a feature with the same name but with the string data type, scoring requests now succeed. The incompatible feature is ignored, scores are returned, and the tango container logs print an error message describing the incompatibility. (PRR-2208)

Upgrade Notes

Use the following upgrade script to have existing models display correctly (as explained in the Resolved Issue entry for PRR-3119, above):

docker exec -it tango /bin/bash
cd src/gpr/common/scripts/versioning/
MODE=prod python3.6 upgrade_44a_predictor_auc_chart.py

Use the following special procedure if you need to upgrade Docker:

  1. Stop AI Core Services (AICS):
    bash scripts/stop.sh
  2. Back up the tango.env file.
  3. Perform the Docker upgrade. See the Docker documentation for the upgrade procedure.
  4. Start Docker:
    systemctl start docker
  5. Enable Docker:
    systemctl enable docker
  6. Re-install AICS:
    bash scripts/install.sh
  7. Merge your backup version of the tango.env file with the one you just installed.
  8. Start AICS:
    bash scripts/start.sh

For HA environments, perform the specified steps on each server running Docker.

This page was last modified on March 26, 2019, at 13:23.


Comment on this article:

blog comments powered by Disqus