Skip to main content

Release Summary - May 02, 2024 (24.09)

The following key features and improvements, along with bug fixes, have been released in Algonomy CXP products in the release version 24.09 during Apr 19, 2024 - May 02, 2024.

Enterprise Dashboard

Enhanced Recommendation Rules for Recently Purchased Categories

We have enhanced our recommendation engine to allow products from recently purchased categories to be recommended. Optimization managers now have the ability to tailor recommendations to align with a shopper's latest purchases, significantly boosting personalization and relevance.

Managers can configure the system to include products from specific categories that a user has recently engaged with, along with the option to set a look-back period to define the recency of eligible purchases. This feature ensures that the recommendations are pertinent and timely, enhancing the shopping experience by suggesting items that reflect the shopper's current interests.

May 02 2024.png

Jira: ENG-27960


Enhanced default setting for Engage CTR optimization

We have updated the default setting for the Engage Click-Through Rate (CTR) optimization model to improve performance. The FTRLv2 model is now the standard option for new sites, ensuring more robust optimization right from the start.

When setting up a new site, the 'Engage Optimization version' in Site Configurations will automatically be set to version 2, aligning with our latest technological advancements for optimal results.

Jira: ENG-16116

Data Engineering

Enhanced Social Proof Output API Response

In our latest update to the Social Proof Output API, we have now integrated the capability to return the rcs (Remote Cookie Store) as part of the API response. This improvement aligns with the behavior of our other personalization APIs, which include rcs in their responses to support seamless integration and continuity across client interactions.

Previously, clients utilizing our new Social Proof API encountered a challenge where the rcs was not accessible on the first call if no prior personalization APIs had been invoked. To address this, we have implemented a solution similar to our recsForPlacement API, where rcs is fetched directly from the rrserver and included in the API output. This update ensures that all server-side clients and apps can immediately leverage the rcs for further personalized actions within the same session, enhancing the overall user experience and consistency across our platform's interactions.

Jira: ENG-28023

Revised Social Proof User Tracking Logic

We have updated the logic for setting the rrUserGuid in the Track Event Experience API calls associated with Dynamic and Social Proof Experiences. Previously, inconsistencies in rrUserGuid implementation caused discrepancies in user identification across different parts of the system. Now, the rrUserGuid is derived directly from RecInternalData, aligning it with the standard computation methods used throughout the platform.

The rcs data, critical for personalized experiences, is now consistently provided in the Social Proof API responses, enhancing the accuracy of user-specific content delivery.

Jira: ENG-27928

Other Feature Enhancements

The following feature enhancements and upgrades have been made in the release version 24.09 during Apr 19, 2024 - May 02, 2024.

Jira #



General Availability




Enable Find response timer logs on Jetty rrServer

To improve performance tracking and diagnostics, we have enabled detailed logging of Find request and response times for specific sites. This feature records the timing data for both ingress and egress, providing valuable insights into the system's operational efficiency and helping identify any potential bottlenecks in real time.





Find, Science:

Default Model Update for Vector Embeddings

We have updated our default model for vector embeddings to the text-embedding-3-large with 1024 dimensions. This change ensures optimal compatibility with Solr9's recommendations and enhances handling of multi-lingual inputs. Now, any request for vector embeddings that does not specify a model will automatically utilize this new default setting.





chatGPT Shopper Assistant - Schedule a job to create vector data for a product catalog

We have streamlined our chatGPT Shopper Assistant by automating vector data generation for product catalogs, enhancing the eCommerce experience. This scheduled job runs daily, updating product vectors and storing them for QA and demonstration. It smartly processes only new or updated products to maintain efficiency.





Streaming Catalog:

Counting view style for enrichment view

We have enhanced the counting functionality in the enrichment view to ensure accurate record tracking. Users can now reliably count the number of items in the view store using updated API endpoints or through specific command-line tools.





Find automatic secondary sorting

We have improved the Find API to automatically include a secondary sorting criterion, ensuring consistent ordering of results when multiple products share the same score. By default, if no secondary sort parameter is provided, products will be sorted by their internal document ID in ascending order after sorting by score.





Prod DLLS:

Update the version of airflow from 1.0 to 2.0.

We have successfully upgraded Airflow to version 2.0 and transitioned the processing environment from the machine "luna" to "dev-inf-051." This update ensures that our workflow management is more robust and efficient, aligning with the latest best practices and capabilities.

The move also involved updating operational scripts, such as those used for log file management, to function seamlessly in the new server environment.


Bug and Support Fixes

The following issues have been fixed in the release version 24.08 during Apr 05, 2024 - Apr 18, 2024.




General Availability



Recommend, rrportal:

CatalogInfo API returns unexpected data for multiple categories

We have addressed an issue where the CatalogInfo API was returning unexpected data for multiple categories. This fix ensures that the API now consistently returns the correct results when querying multiple categories, maintaining data integrity and reliability across the platform.





Composite Outfit - Manually added products not generating outfits

We have resolved a critical issue where manually added products were not generating composite outfits as expected. Previously, when products were manually added to a style with multiple parts, the system failed to create the anticipated outfits.

This fix ensures that outfits are now correctly generated based on the combinations of the manually added products.




Streaming catalog:

GC overhead limit issue for enrichment sidekick

We have successfully resolved a critical 'GC overhead limit exceeded' error in the enrichment sidekick, ensuring that dataset publishing and republishing processes now function efficiently without memory constraints.




Streaming catalog:

Mismatch in catalog and enrichment store item count

We have successfully addressed the issue where there was a discrepancy between the item counts in the catalog and enrichment store. The issue was due to the processing and memory handling of vector items, which has now been optimized to prevent mismatches.




streaming engine not saving all records coming from enrichment

We have resolved the issue where there was a significant mismatch in the counts of enriched items between the enrichment store and the item store, with 3042 items previously unaccounted for. All enriched items are now consistently reflected across both places.


  • Was this article helpful?