pub enum MetricKind {
Show 13 variants
CpuLoad,
MemoryPressure,
Latency,
Coherence,
ErrorRate,
DiskIo,
SwapUsage,
ImaginationQuality,
ResearchOutput,
ScenarioConfidence,
SimulationVariance,
ConformalCoverage,
BrierScore,
}Expand description
Kinds of metrics tracked by the self-model.
Variants§
CpuLoad
CPU load fraction (0.0 = idle, 1.0 = saturated).
MemoryPressure
Memory pressure fraction (0.0 = plenty, 1.0 = critical).
Latency
Dispatch latency in milliseconds.
Coherence
Citta coherence score (0.0–1.0).
ErrorRate
Tool error rate fraction (0.0 = no errors, 1.0 = all errors).
DiskIo
Disk I/O rate fraction (0.0 = idle, 1.0 = saturated).
SwapUsage
Swap usage fraction (0.0 = none, 1.0 = full).
ImaginationQuality
Imagination quality score (0.0–1.0) — scenario evaluation score.
ResearchOutput
Research output rate (0.0–1.0) — hypotheses generated per cycle.
ScenarioConfidence
Scenario confidence (0.0–1.0) — MC rollout positive fraction.
SimulationVariance
Simulation variance (0.0–1.0) — MC std dev (lower is better).
ConformalCoverage
Conformal empirical coverage (0.0–1.0) — fraction of prediction sets/intervals that contained the truth in the latest monitor run. Drift monitoring: sustained drops below the target (1 − α) mean the calibration is stale and should be re-fit.
BrierScore
Average Brier score of resolved predictions (lower is better). Rising Brier = deteriorating calibration honesty.
Implementations§
Source§impl MetricKind
impl MetricKind
Sourcepub const fn higher_is_better(self) -> bool
pub const fn higher_is_better(self) -> bool
Whether higher values are better (true) or worse (false).
Sourcepub const fn default_warning(self) -> f32
pub const fn default_warning(self) -> f32
Default warning threshold for this metric.
Sourcepub const fn default_critical(self) -> f32
pub const fn default_critical(self) -> f32
Default critical threshold for this metric.