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
As the search comparison tool continues to add features that rely on user customization, like #260, the information needed to run search comparisons increases. If a user wants to revisit a previous one, all prompts must be re-entered to achieve the desired comparison.
What solution would you like?
As described in #260, we want consider the concept of a search configuration:
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.
We want users to be able to save search configurations made in the tool. This opens up the option for users to:
Revisit comparisons without re-entering information in the prompts
Share saved comparisons with others using URLs, similar to how saved logs can be accessed in log explorer
What alternatives have you considered?
Currently, the only way to reproduce a search comparison is to copy and paste all the information back into the tool's prompts. As configurations get more complex, this becomes increasingly tedious.
Do you have any additional context?
In preparation for this blog post, saved configurations has a prototype as shown in the ML playground. Users can click links to access saved configurations that have been pre-populated with queries relevant to semantic search.
The text was updated successfully, but these errors were encountered:
Is your feature request related to a problem?
As the search comparison tool continues to add features that rely on user customization, like #260, the information needed to run search comparisons increases. If a user wants to revisit a previous one, all prompts must be re-entered to achieve the desired comparison.
What solution would you like?
As described in #260, we want consider the concept of a search configuration:
We want users to be able to save search configurations made in the tool. This opens up the option for users to:
What alternatives have you considered?
Currently, the only way to reproduce a search comparison is to copy and paste all the information back into the tool's prompts. As configurations get more complex, this becomes increasingly tedious.
Do you have any additional context?
In preparation for this blog post, saved configurations has a prototype as shown in the ML playground. Users can click links to access saved configurations that have been pre-populated with queries relevant to semantic search.
The text was updated successfully, but these errors were encountered: