Release Summary 26.07 | Apr 06, 2026

ENG-31973, ENG-32016

Recommend:

CategoryCP2 Model in Configurable Strategies

The CategoryCP2 (Categories Bought Together) model, based on category co-purchase data, is now available in Configurable Strategies and has been added to the new page under the “Other” model type group. This enables more relevant category recommendations and provides greater control over configuration, including the ability to specify the category tier.

The model supports Category Affinity, Category Context, and Fixed Category Seed as personalization seed options, along with User Affinity, Smart Shuffle, and Margin for sorting. No additional filters are required. Users can enable category recommendations within the strategy and optionally include the seed product in the results.

ENG-32005

Recommend:

Use Primary Category as Seed for Category-Based Strategies

Category-based strategies now use the product’s primary category as the seed when a product seed is provided. This ensures that CategoryTopSellers and related CategoryBestSeller strategies return results that are contextually relevant to the shopper.

This logic also applies to Configurable Strategies that use Category Context as the seed. If a product does not have a primary category defined, the platform continues to follow the existing category selection logic.

ENG-31833, ENG-32015

Recommend:

Product List Model Support in Configurable Strategies

Support has been added for the Product List model (recsUsingCustomInput) in Configurable Strategies and on the new Configurable Strategies page under the “Other” model type group. This enables users to define a fixed list of products for manual recommendations, sort them by affinity, and apply brand or category diversity rules. The enhancement also supports future use in Offer Personalization.

Users can select Product Context or Fixed Product List as seed options. The Fixed Product List supports an unlimited number of products and allows multiple product IDs to be entered at once using a comma-separated format. Sorting options include User Affinity, Smart Shuffle, and Margin. Additional options allow users to diversify by category or brand, enable category recommendations, and include the seed product in the results.

ENG-31582, ENG-31583

New Templates Added to Dynamic Experience UI

Two new templates have been added to the New Dynamic Experience UI: Subscription Footer Bar and Subscription Lightbox. These templates enable users to quickly configure subscription-based experiences in commonly used layout formats.

ENG-32045

Find:

Find Mesos Job Status Report Enhancement

The Find Mesos Job Status Report layout has been updated to improve visibility and clarity. The “JOBS NOT RUN” section is now displayed at the top of the report for quicker identification of issues.

In addition, the report now shows only the latest failed job ID for each failed job, reducing noise and making it easier to track the most recent failure. The enhancement is working as expected.

ENG-31914

Enterprise Dashboard:

New Configurable Strategies Page Enhancements and Fixes

Several updates have been made to improve consistency on the new Configurable Strategies page. The “Advanced Filters” label has been renamed to “Advanced Settings.” The Strategy Preview payload has been aligned with the legacy page, including adding the missing regionExtId, correcting the seed format to a string, and ensuring parameters match the old implementation. For Category or Brand Affinity with Best Seller models, the preview now correctly uses seedStrategy: "FIRST_VALID" unless “Use Multiple Items” is selected.

Additional UI fixes include correcting price labels in the View Results tab, updating the archive icon tooltip to “Archive,” and disabling the “User Wishlist” model when Wishlists are not enabled in Site Config, with a tooltip explaining the restriction.

ENG-32043

Chatbot:

Chatbot Voice Mode Implementation

Voice mode has been implemented in the chatbot to allow users to ask questions and hear responses without typing or reading text. Users can enable or disable voice mode, after which the interaction runs hands-free.

The solution leverages the new voice technology provided by the Science team, enabling automatic detection of when a user has finished speaking and seamless transmission of input to the chatbot. The implementation was completed in collaboration with the Science team and supports smooth entry into and exit from voice mode.

ENG-32024

Find:

SFI: Batching for Category-Based Product Updates

The Streaming Find Indexer (SFI) experienced an Out of Memory (OOM) error during large category updates, where the category fan-out process attempted to update a very large number of associated products in Solr at once. This led to consumer interruption, increased Kafka lag, and stale search data until recovery.

To address this, batching has been implemented in the category product update flow to limit memory usage during large updates. The solution bounds processing volume and improves stability during high-cardinality category events. The enhancement is working as expected.

ENG-31903

Dashboard Chatbot:

Bypass Page Context for Home Page Queries

The chatbot flow has been updated to handle questions asked from the dashboard home page differently. When a user submits a query from the home page, the chatbot no longer attempts to search only within the corresponding documentation page.

Instead, it skips directly to the broader documentation search flow, allowing the LLM to evaluate all available documentation. This prevents limited responses and improves overall response quality and performance.

ENG-31831

Chatbot:

Handling “Show Variety” Question

An issue was identified where the chatbot was unable to answer a question about the “Show variety” option in Advanced Merchandising, even when the context was specified. The root cause was related to configuration and documentation indexing behavior.

Configuration updates were made to improve document loading and embedding retrieval, ensuring the relevant Advanced Merchandising documentation is correctly indexed and accessible. The chatbot can now successfully retrieve and respond with the appropriate explanation. The enhancement is working as expected.

ENG-31832

Helper Chatbot:

Chatbot Prompt Enhancement for Unresolved Questions

The chatbot has been updated to better handle broad or unclear questions. The UI now passes the documentation page name (from Alias.xml) as context, and this value is included in vector embeddings metadata. The LLM first searches within the current page’s documentation before falling back to the broader documentation search.

If no answer is found after both steps, the chatbot now asks the user to provide more context instead of returning a generic response

ENG-32092

Find:

Related Searches Generator Job: Support Multiple Mongo Pipelines

The Related Searches generator job was failing for queryType: "topAndZeroResultSearches" due to an unsupported MongoDB pipeline stage ($facet), resulting in job failure . Since separate pipelines for topSearches and zeroResultSearches already exist, the job has been enhanced to support running multiple Mongo query pipelines instead of relying on $facet.

With this update, the job executes the required pipelines independently and combines the results as needed, avoiding the unsupported stage and ensuring stable execution. The issue has been fixed.

ENG-31834

Enterprise Dashboard:

Fixes to InSegment Option in Configurable Strategies

Updates have been made to improve the InSegment behavior in Configurable Strategies. When custom segments are used in a strategy, the segment dropdown in the View Results tab now correctly displays the segment names.

In addition, when opening a saved strategy, no model is preselected by default, ensuring the configuration reflects the saved state accurately. The issue has been fixed.