Release Summary 25.22 | Oct 31, 2025

ENG-29575

Social Proof:

Social Proof Message-Level Reporting (Rollup)

A new message-level reporting capability has been introduced in Social Proof, allowing merchandisers to measure the performance of individual message types, intervals, and thresholds. The report includes detailed metrics such as sessions, impressions, add-to-cart sessions, purchase sessions, orders, and sales to help analyze which messages drive stronger engagement and conversions.

This enhancement provides greater visibility into message effectiveness, enabling merchandisers to compare and optimize Social Proof messages for improved shopper engagement and conversion performance.

ENG-30910

MVT API:

Support for Multiple Regions and Traffic Availability

The MVT traffic availability API now supports multiple regions, enabling merchandisers to test configurations across selected regions without creating separate tests. When multiple regions are selected, the API calculates and displays the available traffic based on the region with the least available percentage.

This enhancement simplifies multi-region testing, improves efficiency, and ensures accurate traffic distribution insights for better experiment planning and management.

ENG-31087

Recommend:

Expanded Detail View for Rules on Placement Profile Page

The Placement Profile page has been redesigned to provide a clearer and more organized view of rule configurations. Instead of displaying all rule types together, users can now select rule types from tabs displayed along the top of the Rules section. Each tab includes a numeric badge showing how many rules exist for that type. The selected rule type displays its details in a structured table for easier review and management.

Additional enhancements include a multi-option button to add new rules by type and a filter option to view disabled or expired rules. This update simplifies navigation, improves readability, and makes managing placement rules more intuitive.

ENG-31277

Find:

Integration of FileBasedSpellChecker in Search Service

The Search Service now integrates the FileBasedSpellChecker to enhance the accuracy of spellcheck responses. This improvement ensures that user queries are corrected more reliably, providing better search results and reducing the chances of missed matches due to spelling errors.

ENG-31007

Ensemble AI:

Ensemble AI Reporting Enhanced for Style-Level Aggregation

The Ensemble AI Reporting feature has been refined to include only style-level aggregation in the Style Performance Report. Previously, outfit-level aggregation caused inconsistencies where conversion rates appeared as zero in the table view, even though data was reflected in the graph.

With this update, reporting now focuses exclusively on style-level metrics for views, clicks, and purchases. This ensures more accurate and consistent conversion rate calculations across all dimensions, providing clearer insights into performance at the style level.

ENG-30612

Social Proof:

Social Proof Message – Last Purchased X Minutes Ago

The Social Proof API has been enhanced to display how recently a product was purchased or added to cart. New fields for lastPurchase, lastAtc, and lastView timestamps are now included in the API response when the parameter &requireLastEventTime=true is used. This enables real-time messaging such as “A customer bought this 15 minutes ago,” helping drive shopper engagement through timely and relevant cues.

This enhancement provides better visibility into recent shopper activity and strengthens the effectiveness of social proof messaging.
Note: Only API changes have been implemented; UI updates are planned for a future release.

ENG-31165

Integration of FileBasedSpellChecker Configuration with Solr

The Solr configuration now supports FileBasedSpellChecker to improve spell correction and suggestion accuracy. Solr prioritizes entries from the Master Glossary while using its default logic for matching.

For Jaro, HITS, and Suggest modes, Solr now combines glossary-based and default suggestions, ensuring more relevant and consistent search results.

ENG-29227

Find:

Related Searches Configuration Moved to Complementary JSON

The related searches configuration has been moved from the Find index settings to the complementary JSON. This change allows updates to be made at runtime without requiring a new snapshot creation, improving flexibility and reducing configuration overhead.

ENG-31260

Engage:

Campaign Fallback Logic for Multi-Content Placements

The content selection logic has been enhanced to support campaign fallback for placements configured to return multiple contents. When a campaign only partially fills a placement, the engine now continues evaluating additional eligible campaigns in order of priority until the placement is fully populated or no further campaigns are available.

This ensures that placements always display the intended number of contents, maximizing engagement opportunities and maintaining consistent user experiences across campaigns.

ENG-29781

Ensemble AI:

LLM-Generated Ensembles Integrated with Existing Ensembles

Ensemble AI now merges LLM-generated ensembles with existing ones to create richer, AI-driven outfit recommendations. Using text and visual AI, the system identifies complementary products for a seed item and matches them with the best options from the catalog via a Vector Database.

Both Structured and Free Form approaches are supported, allowing flexibility in how ensembles are generated and scored. OpenAI-generated ensembles receive a configurable score boost (default 1.3), ensuring stronger ranking for relevant results. This enhancement delivers more diverse and contextually accurate style combinations for fashion clients.

ENG-30781

Ensemble AI:

LLM Job Enhancements

The Ensemble AI process now automates ensemble generation through improved LLM job handling. The system can generate high-quality ensembles without requiring manual style definitions, using the client prompt and product seed definitions directly from the style configuration.

The update also supports a separately hosted Visual AI model for better scalability and performance. These changes streamline the workflow, reduce manual setup, and help deliver more relevant, visually consistent ensemble recommendations.

ENG-30395

Science:

Recipe API for Seed Products

A new Recipe API allows shoppers to discover recipes related to a specific product, enhancing engagement and increasing average order value. When a seed product is passed, the API returns a list of recipes that include it, along with details such as title, description, image, ingredients, preparation steps, mapped client products, and cooking instructions.

The API also supports dashboard previews by product and locale, and tracks recipe and product clicks for better insight into shopper behavior. This enhancement connects product discovery with practical use cases, helping shoppers see the value behind each recommendation.

ENG-29633

Science:

Recipe – Ingredient to Client Product Mapping

The Recipe feature now maps ingredients from the universal recipe catalog to products in each client’s catalog, enabling shoppers to add ingredients directly to their carts. The mapping identifies a seed product as an ingredient and matches it to related products using catalog enrichment and sales-based ranking.

This process runs daily to stay aligned with catalog updates and shopper affinities, ensuring that the most relevant, top-selling products are linked to each recipe ingredient. If no valid mapping exists, recipes are not displayed for that product. This enhancement creates a seamless bridge between recipe discovery and product purchase.

ENG-30877

Find: 

Normalized Character Accents in Hybrid Search

The hybrid search now normalizes character accents and letter cases when matching query vectors. This ensures consistent results for keywords with or without accents or capitalization differences.

For example, searches like “Tricô” and “Trico” or “Rakhi Dresses” and “rakhi dresses” will now return the same set of results. This enhancement improves search accuracy and delivers a more seamless experience for shoppers using multilingual or mixed-case queries.

ENG-31173

Chatbot:

Incremental Vector Database Updates

The Chatbot and Ensemble AI systems now support incremental updates to the vector database, ensuring that product catalog changes are reflected within an hour. This includes adding new products, updating modified attributes such as brand or category, and removing unavailable items.

When changes occur in the product catalog or attribute configurations, the corresponding updates are automatically applied to the vector database. This improvement keeps conversational and recommendation experiences current, accurate, and aligned with the latest product information.

ENG-30975

Chatbot:

Chatbot History API for Dashboard

The Dashboard now includes API support to retrieve and display chatbot history. A new endpoint returns all past chat sessions for a given user, including the chat title, session ID, and last updated timestamp, allowing users to easily revisit previous interactions.

Another endpoint retrieves the complete conversation for a selected chat, including metadata and messages in markdown format. This enhancement makes it easier to access, review, and continue past chatbot conversations directly from the dashboard.

ENG-31118

Enterprise Dashboard:

Removal of Deprecated Page Types from Placements UI

Deprecated and unused page types have been removed from placement creation, merchandising rules, and Conditional Placements. This update streamlines the dropdown options, ensuring only relevant page types appear and improving overall usability.

ENG-31264

Recommend:

‘Do Not Recommend Recently Viewed Items’ Rule Issue

We have resolved an issue where the “Do Not Recommend Recently Viewed Items” rule was not correctly filtering out previously viewed products. The fix ensures that items recently seen by shoppers are now properly excluded from recommendations across all relevant placements, including Home Page and Mini Bag.

ENG-31121

Recommend:

Conflict Between Boost and Manual Rules in Discover

We have fixed an issue where manually merchandised products were not appearing at the top of Discover results when boost rules were also applied. The system now ensures that manual recommendation rules take precedence over boost rules, displaying manually selected products first as expected.

This fix restores the correct rule hierarchy, ensuring consistent merchandising behavior across environments.

ENG-31216

Recommend:

New Bot Added to Known Bots List for Accurate Analytics

We have verified and added the semrush.com bot to the known bots list to ensure its activity is correctly identified and excluded from analytics. This prevents automated traffic from skewing engagement and performance metrics while maintaining accurate reporting.

ENG-31225

Recommend:

Item Page Rank1 Placement Malfunction

We have resolved an issue where the BrandPopularProducts strategy was inconsistently triggering for certain brands in the item page Rank1 placement. The fix ensures that the strategy now functions reliably across all applicable products, delivering consistent recommendations as expected.

ENG-31287

Chatbot:

Untitled Session Display Issue in Chatbot

We have fixed an issue where a new chatbot session appeared as “Untitled” after all previous sessions were deleted. The chatbot now correctly generates a new session title when a fresh conversation begins, ensuring a consistent and organized chat history.

ENG-31156

Enterprise Dashboard:

iOS Auto-Zoom Behavior on Form Fields

We have fixed an issue where pages were automatically zooming on iOS devices when interacting with form fields. The problem occurred because iOS Safari auto-zooms when text size is below 16px. All form controls now use a minimum font size of 16px, preventing unwanted zooming and ensuring a consistent viewing experience across devices.

ENG-31250

Social Proof:

Social Proof Optimization Experience Playback

We have resolved an issue where the correct experience was not displaying when optimization was enabled. The Social Proof Messages API now functions as expected, ensuring that optimized experiences are properly triggered and shown to users.

ENG-31119

Recommend:

Postgres Connection Pool Error During Checkpointing

We have fixed an issue that caused intermittent “connection is closed” errors during checkpointing. The system now uses a connection pool for the Postgres saver, ensuring stable connections, improved reliability, and smoother checkpoint operations in production.

ENG-31180

Recommend:

Incorrect Domain Links in Configurable Strategies Results Resolved

We have fixed an issue where Configurable Strategies results were returning links with IP addresses instead of the correct domain name. All link URLs, including linkURL, clickTrackingURL, and clickURL, now properly use the domain format, ensuring accurate and consistent URL generation across all strategy results.

ENG-31290

Enterprise Dashboard:

Incorrect ‘DSW Not Activated’ Message Display

We have corrected an issue where the message “DSW has not been activated” was appearing even for sites with DSW already enabled. The message now displays only when DSW is genuinely inactive, ensuring accurate status feedback for all sites.

ENG-31256

Enterprise Dashboard:

Missing Input Fields in Configurable Strategies Preview API Call

We have addressed an issue where input fields, such as the user ID, were not being passed in the recsUsingStrategy API call during Configurable Strategies preview. The API now correctly includes all populated input fields, ensuring accurate and consistent preview results.

ENG-31235

Chatbot:

404 Errors for Chatbot Static Files in Production

We have corrected an issue where chatbot static files were returning 404 errors in production. The context root has been updated to /chatbot/, and the Flask configuration now serves static files from the correct folder, ensuring that all chatbot assets load properly.