I have a question regarding the pricing for Vertex Retail Search.
The billable units are Search Request Units (SRUs), where each unit equals 1,000 requests. Is it correct to assume that one complete search action by a user generates a single search request, or could a search generate multiple requests (e.g., in a “search-as-you-type” setup)?
If search units are affected by a search-as-you-type setup, what are the best practices for debouncing or setting thresholds for minimum characters to balance cost and UX?
In Vertex Retail Search, billing is based on Search Request Units (SRUs), meaning that a single search action can generate multiple requests, particularly in a “search-as-you-type” setup. Each keystroke may trigger a new search request, resulting in multiple SRUs charged for one user action.
Vertex Retail Search pricing is determined by SRUs, with each unit representing 1,000 requests. Normally, one search action generates one request, but several factors can affect the total SRUs consumed:
Search-as-You-Type: In this setup, each keystroke may trigger a new search request, leading to an increase in SRUs.
Faceting and Filtering: Applying facets or filters can also generate additional search requests, impacting the total SRUs used.
To optimize costs and enhance user experience in a search-as-you-type scenario, consider these best practices:
Debouncing: To decrease the number of Search Results Units (SRUs), debouncing can be implemented for search requests. This technique ensures that only the latest query is sent to the server after a brief pause.
Minimum Character Limit: Implement a minimum character count for search queries to reduce unnecessary Search Resource Unit (SRU) consumption for short or incomplete inputs.
Caching: Cache search results to reuse them for future searches, which can also decrease the number of SRUs generated.
For more details on pricing and best practices, you can check these documents:
This is partially incorrect. In regards to caching, the documentation states several times: Warning: Never cache personalized results from an end user, and never return personalized results to a different end user.
##Dear friend, Seehoob, Hello, I analyzed your post and noticed that it is a redirect to the Vertex AI Retail documentation, so did you copy the last line and post it? What is your question regarding the retail product? If you have more information and the context for your question, I can then help you with your query.