@interopio/desktop API Reference

Configuration for the event loop responsiveness instrumentation.

Both measurements are event-driven (no polling): samples record only when the browser delivers a performance entry, so an idle or smooth application produces none. Metric export follows the meter's publish interval as usual.

Properties (10)

Property Type Default Required Description
enabled⚓︎ boolean x

Whether the event loop instrumentation is enabled.
eventLoopBlockingMetric⚓︎ boolean | string | null false x

If true or a string, the main-thread blocking time per long frame (the time beyond the 50ms threshold, per the Long Animation Frames API's blockingDuration; where only Long Tasks are available, task duration minus 50ms) is published as a histogram metric. If a string value is specified, this will be the name of the metric. Otherwise, the default "io.insights.event_loop.blocking.duration" is used (no semantic convention exists for this metric). Zero-blocking frames contribute no sample. The "measurement" attribute says which API produced each sample.
eventLoopBlockingMetricBuckets⚓︎ number[] [0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2.5, 5] x

Explicit histogram bucket boundaries for eventLoopBlockingMetric, in seconds, ascending.
eventLoopBlockingSpans⚓︎ boolean true x

If true, main-thread-blocking long frames are published as spans ("interopio.api.instrumentation.eventLoopBlocking") spanning the frame, with the blocking duration and (where the Long Animation Frames API is available) the attribution of the longest script in the frame.
inputDelayDurationThresholdMs⚓︎ number 16 x

The Event Timing API durationThreshold, in milliseconds: only input events whose total duration (event timestamp until next paint) reaches this value are observed. The browser floor is 16ms.
inputDelayInteractionsOnly⚓︎ boolean true x

If true, only interaction-eligible events (Event Timing entries with a non-zero interactionId - discrete interactions like clicks and key presses, the population INP is computed over) are observed. Disabling it also surfaces continuous events (pointermoves etc.) that meet the duration threshold.
inputDelayMetric⚓︎ boolean | string | null false x

If true or a string, the input event queueing delay (event timestamp until its handlers started running - the event loop latency a user experienced) is published as a histogram metric. If a string value is specified, this will be the name of the metric. Otherwise, the default "io.insights.event_loop.input.delay" is used (no semantic convention exists for this metric).

Only events whose total duration reaches inputDelayDurationThresholdMs are surfaced by the browser, so the distribution describes slow-ish interactions, not all input.
inputDelayMetricBuckets⚓︎ number[] [0.001, 0.0025, 0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1] x

Explicit histogram bucket boundaries for inputDelayMetric, in seconds, ascending. The default brackets the INP "good" (0.2s) and "poor" (0.5s) thresholds.
inputDelaySpans⚓︎ boolean true x

If true, slow input events are published as spans ("interopio.api.instrumentation.inputDelay") spanning the event's timestamp until its next paint, with the input delay / processing / presentation phases and the same element-describing attributes as the clickstream spans (elementIdOrPath, elementSelectorString, ...).
trace⚓︎ boolean true x

If false, no spans are published - metrics-only mode: the input-delay and blocking histograms keep reporting. The master switch over inputDelaySpans/eventLoopBlockingSpans, which refine it per kind.