@Vtrivedy10 How do you detect failures in real tracing data? Let's say if RAG is not performing well, how do you catch that? Don't you need some sort of "expected result" to detect a failure?
How do you build evals from live tracing data? Do you annotate them with help domain experts?
@levelsio Oh no beautiful Europe is getting ruined by the other countries. Again. Those frkn Asian countries always making trouble for us. I don't care what happened during the Industrialism and the entire 1900s, I just care about us beautiful Europeans are the innocent ones.