Loomi Search+ API Controls

This guide covers the process of applying Loomi Search+ via API parameters.

API controls to apply Loomi Search+

To use these parameters, you must have access to the Loomi Search+ feature and already completed vector embedding generation for Loomi Search+ on the dashboard. If these two conditions are not met, default search (keyword-based search) will power the search results.

You can monitor the embedding generation status on the dashboard.

Error handling

If an incorrect value is passed for the above parameters, the system returns a 400 bad request error: “Incorrect value passed for query.search_mode”

AB testing

You can test Loomi Search+ experience vs. Default Search experience by passing the above parameters in the Frontend API call.

Test variant: Enable Loomi Search+ to retrieve the recall

https://core.dxpapi.com/api/v1/core/?
q=dress
&query.search_mode=hybrid
&account_id=6092
&domain_key=homeoasis_bloomreach
&url=www.bloomique.com
&ref_url=www.bloomique.com
&request_type=search
&rows=10
&start=0
&fl=pid%2Ctitle%2Cbrand%2Cprice%2Csale_price
&search_type=keyword

Control variant: Enable Default Search to retrieve the recall

https://core.dxpapi.com/api/v1/core/?
q=dress
&query.search_mode=standard
&account_id=6092
&domain_key=homeoasis_bloomreach
&url=www.bloomique.com
&ref_url=www.bloomique.com
&request_type=search
&rows=10
&start=0
&fl=pid%2Ctitle%2Cbrand%2Cprice%2Csale_price
&search_type=keyword

Note: q=* is not supported for Loomi Search+

Apply vector search precision

Here are the API controls for adjusting vector search precision values for one or all queries. To use these parameters, you must have access to the Loomi Search+ feature and already completed vector embeddings generation for Loomi Search+ on the dashboard. If these two conditions are not met, default search (keyword-based search) will power the search results.

📘

Note

When query.search_mode=hybrid is set, but vector_search.temperature is not set to any value, we will always apply the default vector_search.temperature=standard irrespective of what is configured on the dashboard.

Test Dynamic vector cutoff with external A/B testing tool

Prerequisite: Ensure you've completed the Loomi Search+ setup and the Dynamic vector cutoff configuration in the Bloomreach application. The parameters won't work without both of these in place.

Apply Dynamic vector cutoff (DVC) during A/B test

Use the dynamic_vector_cutoff.enabled parameter to control whether Dynamic vector cutoff (DVC) applies during A/B testing on your external platform.

• To apply DVC in your test, pass: dynamic_vector_cutoff.enabled=true
• To exclude DVC from your test, pass: dynamic_vector_cutoff.enabled=false

Control vector search precision modes

The same vector_search.temperature values reflect the per-query cutoff:

• vector_search.temperature=standard (balanced precision and recall)
• vector_search.temperature=high (higher precision with tighter recall)

DVC uses these settings to choose an appropriate cutoff for each query. The Managed by Bloomreach setting isn't set directly via API.

Error handling

If an incorrect value is passed for the above parameters, the system returns a 400 bad request error: "Incorrect value passed for vector_search.temperature".

AB testing

You can test High vector search precision vs. Standard vector search precision by passing the above-mentioned parameters in the Frontend API call.However, keep in mind that passing “high” Vector search precision can impact the product's discoverability.

Test variant: Apply High vector search precision to retrieve Loomi Search+ recall

https://core.dxpapi.com/api/v1/core/?
q=dress
&query.search_mode=hybrid
&vector_search.temperature=high
&account_id=6092
&domain_key=homeoasis_bloomreach
&url=www.bloomique.com
&ref_url=www.bloomique.com
&request_type=search
&rows=10
&start=0
&fl=pid%2Ctitle%2Cbrand%2Cprice%2Csale_price
&search_type=keyword

Control variant: Apply Standard vector search precision to retrieve Loomi Search+ recall

https://core.dxpapi.com/api/v1/core/?
q=dress
&query.search_mode=hybrid
&vector_search.temperature=standard
&account_id=6092
&domain_key=homeoasis_bloomreach
&url=www.bloomique.com
&ref_url=www.bloomique.com
&request_type=search
&rows=10
&start=0
&fl=pid%2Ctitle%2Cbrand%2Cprice%2Csale_price
&search_type=keyword

📘

Note

When a facet value is interacted with that was introduced by the Vector search component, it might bring in more facet values to the mix as well as some additional products (mostly related) to the recall. This happens because Loomi Search+ considers this interaction as another intent and tries to generate more product discoverability options per user feedback.

Conflict Resolution

Suppose a setting is passed at various levels on the API and Dashboard. Conflict resolution order works as follows:

1. Query Override on Dashboard
2. API parameters passed directly on front-end API call
3. Backend A/B test enabled via Bloomreach
4. Catalog level setting on the Dashboard

Response Details

The response format for Loomi Search + is the same as that for Keyword search.

Debugging

1. Loomi Search+ isn’t getting applied after passing the API parameter. What can be the issue?
   There can be 2 reasons why Loomi Search+ is not getting applied:

The status must show “Off” which means that the embeddings are available, but Loomi Search+ is not applied. If the issue persists, contact your Bloomreach representative.

2. Why am I seeing some irrelevant products after enabling Loomi Search+?

   • Loomi Search+ expands product discovery by combining our strong semantic search with Vector search technology. While this generally improves results, it can occasionally suggest products that aren't directly related to your search term. This happens because Vector search prioritizes product similarity rather than exact keyword matches.
   • To refine results, try increasing the Vector search precision for the query. If you continue to encounter a lot of irrelevant products in your search results, share specific examples with the Bloomreach team. Your feedback is crucial in helping us enhance Loomi Search+ and deliver better results.

3. How do I know which products are brought in the recall via Vector search vs which ones are brought via Keyword search?
   Use Product grid insights to identify the products that the Keyword search and Vector search brought.