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
I would like to be able to attach metadata to user feedback, such as the retrieval result and the text excerpt for that result, the user was commenting on, but also environmental fields related to the application.
Looking at the code, there doesn't seem to be an option to attach metadata to user (API) feedback.
Motivation
Having metadata on feedback means it can be filtered more easily when queried. For example, empowering users to provide feedback on individual retrieval results is absolutely key to developing a solid evaluation test set. This is only possible if we attach some indication of which result and, for convenience, the excerpt they are rating.
Though some features exist in the Langsmith UI, it is way too technical for many non-technical business users - the people we need granular feedback from to refine any LLM system - so being able to simplify this by providing feedback options on the front-end app is very powerful. Without metadata much of this isn't possible and all we can do is capture feedback on the run, which isn't specific enough.
The text was updated successfully, but these errors were encountered:
Feature request
I would like to be able to attach metadata to user feedback, such as the retrieval result and the text excerpt for that result, the user was commenting on, but also environmental fields related to the application.
Looking at the code, there doesn't seem to be an option to attach metadata to user (API) feedback.
Motivation
Having metadata on feedback means it can be filtered more easily when queried. For example, empowering users to provide feedback on individual retrieval results is absolutely key to developing a solid evaluation test set. This is only possible if we attach some indication of which result and, for convenience, the excerpt they are rating.
Though some features exist in the Langsmith UI, it is way too technical for many non-technical business users - the people we need granular feedback from to refine any LLM system - so being able to simplify this by providing feedback options on the front-end app is very powerful. Without metadata much of this isn't possible and all we can do is capture feedback on the run, which isn't specific enough.
The text was updated successfully, but these errors were encountered: