Struct OutcomeEvalset 

pub struct OutcomeEvalset {
    pub hash: String,
    pub field: String,
    pub higher_is_better: bool,
    pub horizons_ms: Vec<i64>,
    pub checkpoints: Vec<Checkpoint>,
    pub baseline: BaselineKind,
    pub min_effect: Option<MinEffect>,
    pub cost: Option<CostBound>,
}

The evalset every LLM-authored, applicable proposal is measured against after apply (docs/loop.md, “Evalset-backed outcomes”). An authored lesson carries no built-in recurrence metric — nothing errors when a lesson is merely useless — so without this the Verify gate has nothing to re-measure for exactly the proposals a human was least able to judge. The host names the evalset and the field; the engine takes the baseline from the newest run journaled BEFORE the proposal and reads the current value from runs journaled AFTER the apply. No baseline run → no metric, never a fabricated one.

Fields

hash: String

The evalset hash (the subject is evalset:<hash> in agent:harness).

field: String

The summary field to read: passed, failed, total, error_rate, or any numeric field the host’s harness writes into the summary.

higher_is_better: bool

Which direction is an improvement — passed and an accuracy are higher-is-better, failed and error_rate are not. Stated by the host because getting it wrong would revert an improvement.

horizons_ms: Vec<i64>

Checkpoints after apply, in ms. Default 1d / 7d / 30d. The older spelling; checkpoints wins when both are given.

checkpoints: Vec<Checkpoint>

The schedule in the deployment’s own unit — {"after_ms": n}, {"after_runs": n} or {"after_grains": n} (a bare integer is ms). A benchmark or CI harness wants [{"after_runs": 1}]: measure at the next graded run after the apply, however soon that is. Empty (the default) means horizons_ms.

baseline: BaselineKind

Which run before the apply the verdict compares against (default newest_before_apply; see BaselineKind for why high_water is opt-in).

min_effect: Option<MinEffect>

The minimum effect size a verdict needs to call a regression (default none: any drop past floating-point slack is a regression).

cost: Option<CostBound>

A cost bound read beside the quality field (default none).

Implementations

impl OutcomeEvalset

pub fn schedule(&self) -> Vec<Checkpoint>

The effective schedule: checkpoints when set, else horizons_ms as time checkpoints.

Trait Implementations

impl Clone for OutcomeEvalset

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for OutcomeEvalset

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

Formats the value using the given formatter.

impl<'de> Deserialize<'de> for OutcomeEvalset

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for OutcomeEvalset

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for OutcomeEvalset

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for OutcomeEvalset