supervision-js
    Preparing search index...

    Interface RenderPreparationArtifactDiagnostics

    Diagnostics for one prepared artifact family.

    interface RenderPreparationArtifactDiagnostics {
        activeFrame?: RenderPreparationActiveFrameDiagnostics | null;
        gateHold?: RenderPreparationGateHoldDiagnostics | null;
        gateHoldCount?: number;
        inFlightCount?: number;
        kind: RenderPreparationArtifactKind;
        precision?: "exact" | "preview";
        maxInFlightCount?: number;
        maxPendingCount?: number;
        maxPreparedCount?: number;
        maxPreparedBytes?: number;
        preparedBytes?: number;
        coarseCount?: number;
        pendingCount: number;
        preparedAheadFrameCount?: number;
        preparedAheadSeconds?: number;
        prefetchCount?: number;
        preparedCount: number;
        refillThresholdCount?: number;
        scheduleBatchSize?: number;
        window?: RenderPreparationArtifactWindowDiagnostics;
    }
    Index
    gateHold?: RenderPreparationGateHoldDiagnostics | null
    gateHoldCount?: number

    Times this family held the playback gate since it opened. Only depth reports it.

    inFlightCount?: number
    precision?: "exact" | "preview"

    A DepthFrame window's depth: "preview" for the preview video, "exact" for exact frames loaded ahead for playback.

    maxInFlightCount?: number
    maxPendingCount?: number
    maxPreparedCount?: number
    maxPreparedBytes?: number

    Retained payload and renderer backing budget, and its current reservation.

    preparedBytes?: number
    coarseCount?: number

    Prepared rasters currently at the coarse tier, owed a fine cook once settled.

    pendingCount: number
    preparedAheadFrameCount?: number

    Frames in the unbroken run of prepared frames starting at the active detection frame.

    This is a count of frames, not of media time. Above 1x the preparer strides, so N frames here do not cover N frame durations of playback, and the two numbers diverge by design. Deriving one from the other is wrong in both directions. With playback gate quality "fine", only fine rasters count as prepared.

    preparedAheadSeconds?: number

    Media-time reach of that same unbroken run, in seconds, measured forward from the active detection frame and wrapping at the media end on a looping clip.

    The run stops at the first frame that is not prepared, so this is how long playback can continue before annotations stop, not the furthest frame prepared anywhere. It is also what a detection gate compares against its required lead, so it stays a single comparable scalar.

    prefetchCount?: number
    preparedCount: number
    refillThresholdCount?: number
    scheduleBatchSize?: number