Error Feed Deep Analysis: On-Demand Trace Investigation
On-demand investigation that runs deeper root cause analysis on an Error Feed issue's trace and produces richer findings than the continuous scan.
About
Every issue in Error Feed comes with analysis generated automatically by the continuous scan: description, root cause, evidence, recommendations. For most issues that’s enough to understand and act on the problem.
Deep Analysis is an additional, on-demand investigation you trigger manually. It runs a more thorough analysis of the issue’s representative trace and produces richer findings, especially around root cause precision and the Recommendations & Fixes section.
When to use
Use Deep Analysis when:
- The continuous scan’s findings feel incomplete or you want more specificity about root cause
- The issue is high-severity and you want confidence in the diagnosis before investing in a fix
- The Overview tab’s Probable Root Cause or Recommendations sections are sparse
- You’re doing a post-incident review and need detailed evidence for a write-up
You don’t need to run it on every issue. Save it for the ones where the standard analysis leaves open questions.
How to run it
Deep Analysis is triggered from the Deep Analysis section in the metadata panel on the right side of the issue detail page.
Open the issue detail page
Navigate to any issue from the Feed list.
Locate the Deep Analysis section in the metadata panel
In the right sidebar, scroll to the Deep Analysis section. If no analysis has been run yet, you’ll see a Run Deep Analysis button.
Click Run Deep Analysis
Click the button. A toast confirms the analysis has started, and you’ll see a progress indicator: “Running analysis…”
Takes about a minute. You can navigate away; analysis continues in the background.
Return to see results
When it finishes, the metadata panel shows “Analysis complete.” Go back to the Overview tab to see the updated findings. Probable Root Cause and Recommendations & Fixes will be populated with more detailed content.
What it produces
Results appear in the Overview tab under Probable Root Cause and Recommendations & Fixes, with more detail than the continuous scan produces.
Specifically, Deep Analysis tends to produce:
- More specific identification of the failing component or step
- More targeted recommendations that reference the actual tool, prompt structure, or workflow pattern involved
- Richer pattern analysis when the cluster has multiple traces with consistent failure characteristics
Re-running analysis
Once Deep Analysis completes, the metadata panel shows a Re-run button alongside “Analysis complete.” Use Re-run when:
- You’ve made changes to your agent and want fresh analysis to confirm whether the root cause has shifted
- The cluster has grown significantly since the last run and you want updated findings
Re-running discards the previous result and generates new findings from scratch against the current representative trace.
Analysis states
| State | What you see | What to do |
|---|---|---|
| Idle | ”Run Deep Analysis” button | Click to start |
| Running | Progress spinner + “Running analysis…” | Wait or navigate away |
| Complete | ”Analysis complete” + Re-run option | Check Overview tab for results |
| Failed | ”Retry Deep Analysis” button | Click to retry |
Note
If no trace is selected or the cluster has zero analyzed traces, the Deep Analysis button is disabled. It re-enables as soon as at least one trace enters the cluster.
Relationship to continuous scan
Continuous scan runs automatically on every sampled trace. Deep Analysis runs once, on demand, against the representative trace. They’re complementary:
- Continuous scan gives you breadth: every trace gets analyzed, issues surface automatically
- Deep Analysis gives you depth: one trace gets a thorough investigation when you need it
Neither replaces the other. Typical pattern: continuous scan surfaces the issue, Deep Analysis helps you understand it well enough to fix it confidently.
Next Steps
Questions & Discussion