Struct PoisoningPolicy 

pub struct PoisoningPolicy {
    pub outlier_threshold: Option<f32>,
}

Configuration for check_for_anomaly.

All knobs are opt-in: the default policy reproduces v0.3.2 behaviour exactly (lexical markers + profile-based heuristics only). Callers attach a policy to the engine via crate::query::MnemoEngine::with_poisoning_policy.

v0.3.3 additions

PoisoningPolicy::with_outlier_threshold enables the embedding-space z-score outlier detector in crate::anomaly::outlier. When set, and when a trained crate::model::embedding_baseline::EmbeddingBaseline exists for the record’s agent, check_for_anomaly adds OUTLIER_SCORE_CONTRIBUTION to the anomaly score whenever the record’s embedding is at least threshold standard deviations from the baseline.

Fields

outlier_threshold: Option<f32>

z-score threshold above which an embedding is considered an outlier. None disables the check. Some(3.0) is a reasonable starting point.

Implementations

impl PoisoningPolicy

pub fn with_outlier_threshold(self, threshold: f32) -> Self

Enable the z-score outlier gate with the supplied threshold.

Example: PoisoningPolicy::default().with_outlier_threshold(3.0)

Trait Implementations

impl Clone for PoisoningPolicy

fn clone(&self) -> PoisoningPolicy

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for PoisoningPolicy

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

Formats the value using the given formatter.

impl Default for PoisoningPolicy

fn default() -> PoisoningPolicy

Returns the “default value” for a type.