Tessary
Tessary duration drift
duration_drift is one of Tessary's built-in classifiers. It watches each call site's latency against its own past, covering turn duration and per-tool duration.
It catches the slowdowns that creep in without an outage: a model that got slower, a tool that started dragging, a prompt change that added seconds to every turn. Comparisons run against the call site's own history, both its recent normal and a reference pinned at the last deploy, so sudden breaks and slow creep both register. A detection reads as a plain multiplicative shift, such as 1.4x slower. Findings go to triage, where the ones ruled real issues open cases.
6 questions
Answered, plainly.
Can duration_drift catch a slowdown that happens gradually, not all at once?Yes. It compares against a rolling recent baseline that catches sudden breaks and one pinned at the last deploy that catches the creep the rolling one absorbs.answer →Can duration_drift tell me which tool call made a turn slow?Yes. It watches turn duration and per-tool duration together, so a slow turn and the specific tool call behind it land in the same finding.answer →Does duration_drift use a fixed latency threshold?No. There's no absolute cutoff; each call site and tool is compared only against its own recent history, since normal speed varies enormously by workload.answer →Is duration_drift turned on by default?Yes. duration_drift is one of three classifiers Tessary calls deterministic and turns on by default, alongside cost_drift and tool_error.answer →What is Tessary's duration_drift classifier?A built-in classifier that watches a call site's turn duration and per-tool duration against its own recent history for a real slowdown.answer →Why did my AI agent get slower even though nothing is throwing errors?A model that got slower on the provider's side, a tool under more load, or a prompt that quietly grew all add latency without ever raising an error.answer →