Take Action on Traces

Documentation Index

Fetch the complete documentation index at: https://arize-ax.mintlify.dev/docs/llms.txt

Use this file to discover all available pages before exploring further.

You’ve explored your traces, spotted agent patterns, and set up evals to flag quality issues automatically. Now you’ve found a problem — a hallucinated response, a slow tool call, a retrieval miss. What next? Arize AX lets you take action directly from the trace detail view, without switching tools or context. Every action starts from the toolbar at the top of a span detail view:

• Fix a prompt — open the span in the Prompt Playground to iterate
• Annotate traces — add human feedback labels and scores
• Assign to a labeling queue — route spans to your team for review
• Curate a dataset from traces — save spans as examples for experiments and fine-tuning

[screenshot: span detail toolbar showing Open in Playground, Annotate, and Add to buttons]

​

Putting it all together

A typical workflow using all four actions:

1. Alert fires — your groundedness monitor detects a quality drop
2. Investigate — filter traces to ungrounded responses, click into a few
3. Annotate — mark a sample as “correct” or “incorrect” to validate the eval
4. Curate — add interesting traces to a test dataset
5. Fix — open a failing trace in the Playground, iterate on the prompt
6. Validate — run an experiment to confirm the fix
7. Queue — send remaining traces to a labeling queue for team review

---

​

Next step

You’ve got the core loop down: see traces, evaluate quality, fix problems. Now set up deeper observability — starting with agent visualization: