Struct AbTest 

pub struct AbTest {
    pub id: String,
    pub variants: Vec<Variant>,
    pub assignments: RwLock<HashMap<String, String>>,
    pub metrics: HashMap<String, Arc<VariantMetrics>>,
    pub start_time: Instant,
    pub traffic_pct: f64,
}

An active A/B test comparing two or more variants.

Fields

id: String

Unique test identifier.

variants: Vec<Variant>

All variants in this test.

assignments: RwLock<HashMap<String, String>>

Map of session_id → variant_id for all assigned sessions.

metrics: HashMap<String, Arc<VariantMetrics>>

Per-variant metrics (Arc so VariantMetrics with atomics can live behind shared refs).

start_time: Instant

When the test was started.

traffic_pct: f64

Fraction of traffic routed into this test (0.0–1.0).

Implementations

impl AbTest

pub fn new( id: impl Into<String>, variants: Vec<Variant>, traffic_pct: f64, ) -> Self

Create a new A/B test.

• id — unique test name
• variants — must be non-empty; weights are normalised internally
• traffic_pct — fraction of sessions to route into this test (0.0–1.0)

pub fn assign(&self, session_id: &str) -> Option<&Variant>

Deterministically assign a session to a variant using consistent hashing.

Returns None if traffic sampling excludes this session or there are no variants.

pub fn record_result( &self, session_id: &str, latency_ms: u64, tokens: u64, success: bool, )

Record the outcome of a request for the session’s assigned variant.

pub fn winner(&self) -> Option<&str>

Return the variant_id of the winner if its success_rate exceeds all others by more than 5 percentage points.

pub fn summary(&self) -> String

Render a markdown table summarising per-variant statistics.