Experiments (A/B tests)

getExperiment(name, defaultParams) assigns the user to a variant and returns an ExperimentResult.

ExperimentResult

data class ExperimentResult(
    val inExperiment: Boolean,   // is the user enrolled?
    val group: String,           // assigned group/variant name
    val params: Any?,            // variant parameters (your typed payload)
)

Bound Client form

val flags = Client(currentUser)

val r = flags.getExperiment("checkout_button", mapOf("color" to "blue"))
if (r.inExperiment) {
    @Suppress("UNCHECKED_CAST")
    val params = r.params as? Map<String, Any?> ?: emptyMap()
    val color = params["color"]   // variant param
}

The second argument is defaultParams: returned as params when the experiment yields no params (e.g. not enrolled / not loaded). Pass null for none.

Tracking conversions — track(...)

Record a conversion / success event so the experiment can compute lift. Use the bound Client you already hold from getExperiment — the unit is derived from the bound attribute bag (user_id, else anonymous_id), so there is no user argument:

val flags = Client(currentUser)

val r = flags.getExperiment("checkout_button", mapOf("color" to "blue"))
// …present the treatment…
flags.track("{{SUCCESS_EVENT}}", mapOf("amount" to 49))

• event — the metric event, e.g. {{SUCCESS_EVENT}}.
• props — optional event props (private attributes are stripped, see Advanced).

track is fire-and-forget and never throws into the request path. It is a no-op when the bound bag carries no unit, and under configureForTesting / configureForOffline.

Manual exposure — logExposure(...)

The server is stateless and never auto-logs exposures. Call logExposure on the bound Client at the point you actually present the treatment so the experiment counts the exposure (no user argument — the unit comes from the bound bag):

flags.logExposure("checkout_button")

It re-evaluates the experiment for the bound user and POSTs a single exposure event only if the user is enrolled (no-op otherwise, and in test mode).