Bug: Evaluator feedback metrics show IDs instead of names in observability table

[mmabrouk]

Description

In the main observability table, evaluator feedback metrics are displayed with their IDs (UUIDs) in addition to the human-readable metric names.

Expected Behavior

Metric names should be shown (e.g., 'accuracy', 'relevance') instead of UUIDs.

Actual Behavior

IDs/UUIDs are displayed, making it hard to understand which metric is which.

Location

Main observability table (the table listing all traces)

Severity

Low — cosmetic issue

Steps to Reproduce

1. Run an evaluation that generates feedback metrics
2. Go to the observability table
3. Look at the feedback/metrics columns
4. metric name + IDs are shown instead of metric names

[linear]

AGE-3610

[dosubot]

This is a known UI issue: the observability table currently shows metric UUIDs instead of human-readable names because the data pipeline passes metric IDs straight from the backend to the table without mapping them to names. The mapping should happen during data transformation—ideally when aggregating annotations or right before rendering—by joining metric IDs with evaluator metadata (which includes human-readable names like "accuracy" or "relevance") source.

There’s an open ticket tracking this bug (AGE-3426), and similar display issues have been discussed for other UI components source. No recent PRs have fixed this yet.

To resolve it, update the transformation logic (for example, in attachAnnotationsToTraces or the table render function) to replace metric UUIDs with their corresponding names using evaluator definitions or annotation references. If the mapping isn’t available in the frontend, you may need to fetch evaluator metadata via API or update the backend to include both ID and name for each metric in the response.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}