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
With the launch of search pipelines, we want users to be able to select a pipeline for two reasons: 1) ease of use/clarity, 2) certain processors cannot be used in an ad-hoc search pipeline because the initialization is too costly, especially when connecting to external services.
What solution would you like?
A user of search pipelines should be able to compare two different queries using the same or different search pipelines. We also should consider a solution that makes room for search configurations. Think of a search configuration as a unit containing anything that you would want to compare to another search configuration for manual comparison or A/B testing. This would include queries, ML models, search pipelines, processor configurations, analyzers, etc. In this case we really want to include only search pipelines.
What alternatives have you considered?
Comparing two different sets of results with different pipelines is possible using two Dev Tools windows or two curl commands. These are not as user friendly.
@kgcreative thinking about just adding the search pipeline selector to the tool, but there may be a better way to grow into search configurations... Thoughts?
Is your feature request related to a problem?
With the launch of search pipelines, we want users to be able to select a pipeline for two reasons: 1) ease of use/clarity, 2) certain processors cannot be used in an ad-hoc search pipeline because the initialization is too costly, especially when connecting to external services.
What solution would you like?
A user of search pipelines should be able to compare two different queries using the same or different search pipelines. We also should consider a solution that makes room for search configurations. Think of a search configuration as a unit containing anything that you would want to compare to another search configuration for manual comparison or A/B testing. This would include queries, ML models, search pipelines, processor configurations, analyzers, etc. In this case we really want to include only search pipelines.
What alternatives have you considered?
Comparing two different sets of results with different pipelines is possible using two Dev Tools windows or two curl commands. These are not as user friendly.
Do you have any additional context?
There is a prototype available here for search pipeline selection: https://ml.playground.opensearch.org/app/searchRelevance#/
The text was updated successfully, but these errors were encountered: