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
Although the langsmith experiment viewing interface is pretty good for basic usecases, for my (quite complicated) usecase I would like to build a custom offline experiment comparison tool (with e.g. streamlit or dash or similar). However I do not see a method in the sdk to download a given experiment run from the platform?
Motivation
In order to do error analysis on runs with complicated nested structured outputs from images, I need an interface that allows me to quickly compare specific metrics for each extracted field and e.g. highlight them in the source pdf, etc. I can build this myself, but need access to the underlying experiment result.
The text was updated successfully, but these errors were encountered:
Agreed that a dedicated endpoint (and button in the UI) would be good.
The dataframe format is supported in the SDK (but it's really just a composition of endpoints rather than a dedicated endpoint) - that PR adds the abiltiy to export as json/other formats.
Feature request
Although the langsmith experiment viewing interface is pretty good for basic usecases, for my (quite complicated) usecase I would like to build a custom offline experiment comparison tool (with e.g. streamlit or dash or similar). However I do not see a method in the sdk to download a given experiment run from the platform?
Motivation
In order to do error analysis on runs with complicated nested structured outputs from images, I need an interface that allows me to quickly compare specific metrics for each extracted field and e.g. highlight them in the source pdf, etc. I can build this myself, but need access to the underlying experiment result.
The text was updated successfully, but these errors were encountered: