Struct EvalRun 

pub struct EvalRun {
    pub run_id: String,
    pub passed: u64,
    pub failed: u64,
    pub summary: Map<String, Value>,
    pub recorded_ms: i64,
    pub spend: Option<RunSpend>,
}

One recorded execution of an evalset.

Fields

run_id: String

The eval- run id the cases were journaled under.

passed: u64

failed: u64

summary: Map<String, Value>

The whole summary object, so host-defined fields (category_accuracy, …) are reachable without this module having to know them.

recorded_ms: i64

When the summary grain was written.

spend: Option<RunSpend>

What the run spent, when the runtime journaled it: the run_outcome Observation areev run writes for the same run id. Read only for a cost key the summary does not carry, so an eval run that IS an areev run run quotes one spend to both the Verify gate and the run_outcome analyzer.

Implementations

impl EvalRun

pub fn field(&self, name: &str) -> Option<f64>

A numeric field of the summary. passed/failed are promoted to typed fields but stay readable here too, so a metric string may name any of them uniformly.

pub fn total(&self) -> u64

Cases that ran. 0 when the summary records neither.

pub fn effects(&self) -> Option<u64>

Tool calls the cases made. Summary key effects only — the runtime’s Observation counts supersteps, not effects, so there is no fallback.

pub fn tokens(&self) -> Option<u64>

Input + output tokens; both keys must be integers.

pub fn usd_micros(&self) -> Option<u64>

pub fn wall_ms(&self) -> Option<u64>

Trait Implementations

impl Clone for EvalRun

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 EvalRun

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

Formats the value using the given formatter.