LLM Insights on Android

Platform-specific usage for the Android SDK. For what LLM Insights are, the request/response model, required metrics, and the privacy notice, see the LLM Insights overview.

The insight API lives on the SmartSpectraSdk singleton (SmartSpectraSdk.shared).

Insights summarize the buffered vitals, so breathing (the default set) and cardio must both be active. Assign both bundles to requestedMetrics:

sdk.config.requestedMetrics = buildList {
    addAll(SmartSpectraConfig.breathingMetrics)
    addAll(SmartSpectraConfig.cardioMetrics)
}

cardioMetrics includes MetricType.ARTERIAL_PRESSURE_TRACE, which drives the on-screen pulse waveform. When requestedMetrics is unset the SDK measures breathing only.

Insights are delivered through a LiveData, observed on the main thread. It carries both the auto-fired periodic vitals insights and on-demand responses:

public val insight: LiveData<Insight?>

sdk.insight.observe(viewLifecycleOwner) { insight ->
    if (insight == null) return@observe
    if (insight.requestId != pendingRequestId) return@observe // correlate the reply
    val text = when {
        insight.hasAnalysis() -> insight.analysis
        else -> getString(R.string.insights_chat_error)
    }
    showMessage(text)
}

Match insight.requestId against the value returned by requestInsight to tell on-demand replies apart from auto-fired vitals (which won't match a request you made). Every insight is currently delivered with type == INSIGHT_TYPE_VITALS, so correlate on requestId, not type.

Call requestInsight on a running session. It returns the request ID used to correlate the asynchronous response:

public fun requestInsight(text: String): Int

pendingRequestId = runCatching { sdk.requestInsight("Summarize my current vital signs and flag anything unusual.") }
    .onFailure { /* dispatch failed */ }
    .getOrNull()

The prompt is combined with the latest buffered metrics when they exist, otherwise sent prompt-only.

Insight is the generated proto type (com.presagetech.smartspectra.proto.InsightsProto.Insight). Branch on insight.hasAnalysis() / insight.hasError() (exactly one is set), read the text with insight.analysis / insight.error, and correlate with insight.requestId. (Every insight is currently typed INSIGHT_TYPE_VITALS, so use requestId, not type, to distinguish replies.) Full field documentation is in Data Types → Insight.

The first auto-fired insight arrives about 15 seconds after the session starts; allow that much valid measurement before an on-demand request can be grounded in the user's physiology. The service may return no insight for a given request, in which case the observer is not called.

• LLM Insights overview
• Android API reference
• Data Types