Enum Verdict 

pub enum Verdict {
    Agreed,
    NearMiss {
        rank: usize,
        score: f32,
    },
    Buried {
        rank: usize,
        score: f32,
    },
    Absent,
    NoRanking,
}

How ski’s ranking lines up with the native chooser’s pick on one prompt.

Variants

Agreed

ski injected the same skill the model then used. Redundant — and because the model saw ski’s nudge first, this does not prove ski caused the pick.

NearMiss

ski abstained but ranked the native pick in its top [NEAR_MISS_RANK] — the tunable edge: a looser gate or stronger channel could have surfaced it.

Fields

rank: usize

score: f32

Buried

ski abstained and ranked the native pick deeper in its considered top-K.

Fields

rank: usize

score: f32

Absent

ski never surfaced the native pick in its considered top-K — the retrieval ceiling; narrowing the gate cannot win this one.

NoRanking

No considered ranking was logged for the prompt (telemetry was off then, or a pre-feature event), so the pick can’t be placed.

Trait Implementations

impl Debug for Verdict

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl PartialEq for Verdict

fn eq(&self, other: &Verdict) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl StructuralPartialEq for Verdict