Jump to: navigation, search

9.0.014.00

AI Core Services Release Notes


AI Core Services is part of 9.x starting in 9.0.006.05.
Release Date Release Type Restrictions AIX Linux Solaris Windows
12/21/18 General Under Shipping Control X

Helpful Links

What's New

This release includes the following new features and enhancements:

  • You can now upload Dataset, Agent Profile, and Customer Profile data to Genesys Predictive Routing (GPR) from CSV files that use certain legacy encodings (listed below). By default, GPR always assumes the CSV files are encoded with UTF-8. This change applies to uploads using both the GPR web application and the GPR API. The following encodings are supported:
    • UTF-8
    • Shift JIS
    All data returned from GPR uses UTF-8 encoding.
    (PRR-3826; PRR-3809)
  • GPR has optimized how cardinalities are stored. Cardinalities are now written into a dedicated database collection, so that the cardinalities for each field are stored in their own document. Previously, the cardinalities were stored along with the schema data. With high-cardinality features, this could lead to performance degradation due to additional conversions needed to extract the cardinality data. (PRR-3769)
  • This release improves the performance of the Create Predictor functionality in the GPR web application. Previously, especially with Datasets containing a large number of high-cardinality features, it could take up to 15 seconds to display all the attributes that should appear in the Agent ID drop-down menu. (PRR-3745)
  • The GPR API has improved how it handles certain requests involving Dataset start and end times. Specifically, this addresses the following issues:
    • A GET request to the include_data_distribution endpoint now returns fields that specify the Dataset start (from_dt)and end (to_dt) dates.
    • The API sets the start and end dates for a Predictor Dataset to the actual Dataset boundaries if those dates were not passed explicitly in a request. Previously, if you did not specify the start and end dates, GPR took the dates from 1970. (PRR-3635)
  • The schema management workflow for Agent and Customer Profiles has been simplified and streamlined. The Discovered Fields tab has been removed and cardinality counts have been added to the schema view. This change ensures GPR always presents up-to-date Profile information. The schema tab always presented updated information, if available, but the Discovered Fields tab display was generated only once and did not reflect changes to the Profile schema. (PRR-2886)
  • AICS now supports deployment in an environment running on a Kubernetes cluster. For pre-requisites and deployment instructions, see (Optional) Installing AICS on a Kubernetes Cluster.

Resolved Issues

This release contains the following resolved issues:


You can now override the value set in the GPR web application for the Actions Cutoff parameter in the Predictors configuration window using the GPR API. Previously, if the Actions Cutoff parameter was less than the value sent via API, or if it wasn't set at all, you could not override it in a scoring request from the API. (PRR-3912)


If user does not have values set in the First Name and Last Name fields, the GPR web application now displays the user email address in the upper navigation bar. Previously in this scenario, the GPR web application displayed None None in the top navigation bar. (PRR-3837)


The Inter-Agent Variance graph now displays correctly after you switch from report to report. Previously, when moving from one report to another several times, the graph sometimes rendered incorrectly or not at all. (PRR-3797)


The documentation explaining how to create and understand Agent Variance reports has been clarified and expanded. (PRR-3791)


If you create and generate a Predictor and then try to make an update to it, such as adding a new Agent or Customer feature, you now correctly receive an error message instructing you to purge the existing data before making the update. Previously in this scenario, the upload process stalled without generating an error message or allowing the user to continue. (PRR-3772)


The GPR web application now displays the name of the file last used to append data to a Dataset. Previously, GPR always displayed the name of the file used to create the Dataset. (PRR-3730)


When creating a Predictor from a Dataset with many high-cardinality features, GPR now returns only data relevant to Predictor creation. Previously in this scenario, some parts of the display, such as the date-range histogram chart, were not correctly rendered in the GPR web application. (PRR-3721)


The GPR web application now consistently displays cardinalities for Agent and Customer Profiles and Datasets. In all cases, the exact cardinality now appears for all values through 1001. Any fields with cardinalities larger than 1001 also display 1001. (PRR-3597)


The API no longer allows users to append data to an out-of-sync Dataset (a Dataset for which the schema was not saved and accepted). Previously, this behavior could lead to a broken Dataset if the incoming data was not restricted by a schema. Now in this scenario, the GPR API returns the following error message: {"error": "Cannot append data as dataset is OUT_OF_SYNC"}. (PRR-3447)


The way the GPR API Agent and Customer Profile endpoints handle DELETE requests when nothing is deleted has been corrected so that they always return {removed_count:0} if no record was deleted. For example, an incorrect Agent ID might be passed in the request, so that no record is identified for deletion. Previously in such scenarios, the API would return a non-zero value for the removed_count parameter. (PRR-2730)


Both the GPR web application and the GPR API now accept data uploads that include Agent Groups and Skills with spaces in the names. (PRR-2110)


When you copy a Model, the new Model is created with the same train/test split as the original. Previously, the copy had the train/test split set to the default, which is 80/20. (PRR-1200)


Upgrade Notes

Use the following special procedure to upgrade to release 9.0.014.00.

To perform the upgrade, run the following commands:

docker exec -it tango /bin/bash
cd src/gpr/common/scripts/versioning/
MODE=prod python3.6 upgrade_42a_coll_based_cardinalities.py
exit
cd <IP>
bash scripts/restart.sh

If you need to roll back your upgrade and return to your previous version of AI Core Services, run the following commands:

docker exec -it tango /bin/bash
cd src/gpr/common/scripts/versioning/
MODE=prod python3.6 upgrade_42a_coll_based_cardinalities.py --down
exit

After running these commands, follow the instructions in the Deployment and Operations Guide to install the desired older version.

This page was last edited on December 21, 2018, at 23:18.
Comments or questions about this documentation? Contact us for support!