Find a pattern
Go to Patterns and select a pattern from the list to open it. The list shows open patterns by default:- Click Filter patterns to choose Open, Closed as resolved, Closed as not useful, or Closed without reason. Each option shows its pattern count.
- Use the search box above the list to search by name.
Read a pattern
Check Loop’s reasoning against the evidence before acting. Each pattern has four parts:- Summary: The finding and its context.
- Suggested fix: A recommended investigation or change, when available.
- How to monitor: A recommended scorer, classifier, or facet, when available. See Act on a pattern.
- Evidence: The measurements and traces supporting the finding.
- Frequency measurements when trace data can reliably identify the behavior.
- Monitors for complementary measurements such as latency, cost, or score trends.
- Traces with Loop’s notes quoted inline. Select one to open it in the Debugger, scrolled to the source span field with quoted terms highlighted.
Frequency
The Frequency chart shows the percentage and number of affected traces. Each trace counts once, even if several spans match. Use the selector above the chart to change the time range. When Loop identifies a comparable population, you can choose the scope:- Similar (default): The share of comparable traces affected, such as requests for the same operation.
- All: The share across your project.
Act on a pattern
Choose an action based on whether you want to investigate, track, fix, or share the finding.Keep investigating
Select Continue in Loop to start a new thread with the pattern attached as context. A badge shows the pattern’s name.Make it measurable
Under How to monitor, click Create scorer or Create classifier, depending on Loop’s recommendation. Loop opens with a prefilled prompt, adds the evaluator to the pattern, plans a backfill with you, and creates a monitor chart. This tracks the behavior continuously at lower cost than repeated investigation. Loop can also create a facet so Topics classifies the behavior across all your traces.Hand it to a coding agent
Click Copy pattern as prompt in the pattern header. Braintrust copies a Markdown brief containing the pattern’s status and close outcome, summary, suggested fix, evidence trace IDs, andbt CLI commands for pulling the full traces. Paste it into your coding agent to start work on a fix.
Share it
When the pattern is open in Loop’s workspace, click Copy link in its header. The link only works for people who already have project access.Close or reopen
Closing and reopening requires the project’sUpdate permission, granted to Owners and Engineers permission groups by default. Without it, you can read patterns and evidence, but the Close and Reopen buttons don’t appear.
Close a pattern when you’ve acted on it or decided it isn’t useful:
- Click Close in the pattern header.
- Optionally choose Resolved if the problem is fixed, addressed, or no longer happening, or Not useful if it’s noise, transient, or unimportant.
- Add optional feedback, then click the close button below the feedback field.
How Loop uses closure feedback
How Loop uses closure feedback
Loop uses the outcome and your feedback when evaluating related findings:
- Not useful tells Loop not to automatically reopen or recreate a matching pattern while your reason still applies.
- Resolved, or closing without an outcome, allows Loop to reopen the existing pattern if the behavior recurs, unless your feedback says otherwise.
- Free-text feedback defines the scope of your judgment. If you dismiss synthetic staging traffic but new evidence comes from production, Loop evaluates that evidence independently. Be specific about what you want to dismiss and any conditions for reopening.
When Loop closes patterns automatically
When Loop closes patterns automatically
Scheduled runs also review active patterns. Loop can close a pattern when recent, relevant data shows that the behavior has stopped. Missing data or a failed query is not evidence that it has stopped.
Next steps
- Write scorers to measure a behavior a pattern surfaced.
- Build datasets from the traces behind a pattern, then run an experiment to test a fix.
- Create custom facets to track the behavior across every trace.
- Debug a trace to dig into a single piece of evidence.