Overview

The Overview answers “is everything fine?” over a window of time: the last hour, 24 hours, 7 or 30 days. Every figure, bar and row opens the runs behind it in Observability.

The Overview: figures, runs by outcome over time, what fails most, failing tools

Figure	Shows
Runs started	Runs started in the window, against the previous window
Completed	The share of the runs that ended which completed
Failed	Failed runs, plus the aborted and cancelled ones
Waiting now	Active runs waiting for the outside world, and how many wait on a signal
Duration p95	The 95th and 50th percentiles of the duration of ended runs
Quarantined	Runs held until an operator releases them or gives them up

Below them:

Runs started, by outcome: completed, failed and still active, over the window.
What fails most: failed runs grouped by fingerprint, the error with its numbers, ids and strings replaced, so the same failure across many different scripts is one row.
Failing tools: the tools whose calls failed in the window.

The query bar at the top narrows everything to a query, such as @agent:triage or namespace:user_ada. Save keeps it, and the query and the window follow you into Observability.

Get started

Runs

Code, tools and end users

Your organization