The public status page
Measured success rate and latency for every engine we sample, published openly at /status.
What it is
A public page reporting, per engine, the share of sampled runs that completed over the last 30 days, how many samples that is, median and 95th-percentile latency for the ones that succeeded, and the date of the last success.
When to use it
When a measured run fails or a number looks off, and you want to know whether the engine behind it is healthy before you go hunting through your own site.
Step by step
- 1Open /status. No account needed.
- 2Find the engine your measurement depends on.
- 3Read its state (operational, degraded, failing, or no recent runs), its success rate, and how many samples that rate is computed over.
- 4Compare the latency columns when runs are slow rather than failing.
What you will see
A row per engine we sample with its state, success rate, sample count, median and 95th-percentile latency, and last success date. The figures refresh every few minutes and aggregate runs across the whole platform, carrying no customer, asset, or prompt.
Measured, never manufactured
A failed sample is excluded from your scores and counted here instead, never shown to you as a zero. Latency percentiles cover successful samples only, because a failed request is not a measure of how fast an engine answers. And with no runs in the window the page says so and shows nothing, rather than printing a fabricated uptime figure, which is the same rule every other number in the product follows. A tool that sells measurement should be willing to publish whether its own instruments are working.
Related guides
Next
What a measured run costsBefore you trigger a paid measurement, we show what the last runs of that action actually cost on your site.