You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
No. This is definitely a "nice to have" for what I think is mostly not a problem.
What solution would you like?
If I'm running the search comparison tool to tune the relevance of returned search results, I put a baseline (current production) query in the left-hand pane and my "experimental" queries in the right-hand pane. As I iterate, I'm modifying the experimental query, but I don't touch the baseline. Nonetheless, the Search Relevance UI sends both requests to OpenSearch, when the baseline results (probably) haven't changed.
While the added traffic is pretty small (since I, as a slow human, am not clicking that submit button very frequently), I would like my OpenSearch cluster to do as little work as possible. Also, I recognize that search results may change as a result of updates to the index, which may make the comparison against "old" results invalid. Still, in cases where I know that the index doesn't receive updates, I would like to be able to cache old results with a configurable TTL.
What alternatives have you considered?
Another option might involve explicitly letting the user run only the left or right query in addition to running both. That may be less confusing than a configured TTL, but might clutter the UI. (I'm imagining the result caching + TTL could be in a separate "settings" modal with a check mark for result caching that enables a TTL text box with up/down arrows or something.)
Do you have any additional context?
No
The text was updated successfully, but these errors were encountered:
@reta pointed out that there are non-query factors (like search pipelines, especially if they call out to an external service) that can impact the search results, such that the same query could have very different results.
Is your feature request related to a problem?
No. This is definitely a "nice to have" for what I think is mostly not a problem.
What solution would you like?
If I'm running the search comparison tool to tune the relevance of returned search results, I put a baseline (current production) query in the left-hand pane and my "experimental" queries in the right-hand pane. As I iterate, I'm modifying the experimental query, but I don't touch the baseline. Nonetheless, the Search Relevance UI sends both requests to OpenSearch, when the baseline results (probably) haven't changed.
While the added traffic is pretty small (since I, as a slow human, am not clicking that submit button very frequently), I would like my OpenSearch cluster to do as little work as possible. Also, I recognize that search results may change as a result of updates to the index, which may make the comparison against "old" results invalid. Still, in cases where I know that the index doesn't receive updates, I would like to be able to cache old results with a configurable TTL.
What alternatives have you considered?
Another option might involve explicitly letting the user run only the left or right query in addition to running both. That may be less confusing than a configured TTL, but might clutter the UI. (I'm imagining the result caching + TTL could be in a separate "settings" modal with a check mark for result caching that enables a TTL text box with up/down arrows or something.)
Do you have any additional context?
No
The text was updated successfully, but these errors were encountered: