Struct HeuristicEntry 

pub struct HeuristicEntry {

    pub motif_class: MotifClass,
    pub reason_code: ReasonCode,
    pub candidate_interpretation: &'static str,
    pub provenance: Provenance,
    pub recommended_action: PolicyState,
    pub drift_threshold: f64,
    pub slew_threshold: f64,
    pub boundary_density_threshold: f64,
    pub min_correlation_count: u16,
    pub max_correlation_count: u16,
    pub min_duration_windows: u16,
    pub max_duration_windows: u16,
    pub weight_drift: f64,
    pub weight_slew: f64,
    pub weight_boundary: f64,
    pub weight_correlation: f64,
    pub weight_duration: f64,
    pub evidence_dataset: &'static str,
    pub evidence_dataset_doi: &'static str,
    pub dashboard_hint: &'static str,
    pub taxonomy_ref: &'static str,
    pub affinity_tiers: u32,
    pub confuser_motif: Option<MotifClass>,
    pub margin_vs_confuser_threshold: f64,
    pub primary_witness_tiers: u32,
    pub primary_witness_detectors: &'static [&'static str],
}

Heuristics bank entry — typed, provenance-aware motif record.

The struct is Copy + Clone + Debug + PartialEq; all fields are stack-only ('static string slices, primitives, enums). The bank is initialised at compile time as an array of HeuristicEntry, so every field must be const-friendly.

The seven original fields (motif_class … slew_threshold) preserve the v0.1 wire shape. The thirteen additional fields enable episode-level multi-feature scoring (match_episode), per-motif provenance ladders (evidence_dataset, evidence_dataset_doi), production-engineer dashboard hints (dashboard_hint), and taxonomy anchors (taxonomy_ref). All are documented in docs/heuristics_bank.md.

Fields

motif_class: MotifClass

reason_code: ReasonCode

candidate_interpretation: &'static str

NOT an attribution — a candidate interpretation hypothesis

provenance: Provenance

recommended_action: PolicyState

drift_threshold: f64

Minimum drift persistence (fraction of window) to trigger

slew_threshold: f64

Minimum slew magnitude to trigger

boundary_density_threshold: f64

Minimum boundary-density (fraction of episode windows in Boundary state) for this motif to trigger.

min_correlation_count: u16

Minimum number of contributing signals (e.g. multi-service motifs require ≥ 2). Use 1 for “any”.

max_correlation_count: u16

Maximum number of contributing signals (e.g. single-service motifs require ≤ 1). Use u16::MAX for “unbounded”.

min_duration_windows: u16

Minimum episode duration in windows.

max_duration_windows: u16

Maximum episode duration in windows. Use u16::MAX for “unbounded”.

weight_drift: f64

Per-motif scoring weights for match_episode. Default 1.0 reproduces the v0.1 unit-weighted behaviour; differential weights let one motif emphasise drift while another emphasises slew or correlation.

weight_slew: f64

weight_boundary: f64

weight_correlation: f64

weight_duration: f64

evidence_dataset: &'static str

Provenance ladder — which named upstream slice motivated this entry. "FrameworkDesign" for hand-coded entries with no dataset evidence yet; "<dataset_key>" (e.g. "tadbench_F04", "aiops_challenge_packet_loss") for entries observed in a vendored real-data fixture.

evidence_dataset_doi: &'static str

DOI of the upstream archive cited in evidence_dataset. Empty string when evidence_dataset == "FrameworkDesign".

dashboard_hint: &'static str

One-line hint for a production debug engineer reading the matched motif on a dashboard. E.g. `“Inspect jvm.memory.heap.used

• gc.duration over the past hour“`.

taxonomy_ref: &'static str

Taxonomy anchor — IEEE 24765 term and Avizienis–Laprie–Randell node. E.g. "IEEE 24765: 'fault propagation'; A-L-R: error → service-failure".

affinity_tiers: u32

Phase 2.5 — hand-curated tier-affinity bitmask. Each bit corresponds to one detector tier (see TIER_BIT_* constants in heuristics_bank.rs). The bank’s match_episode_with_tier_affinity AND-s this mask against the per-cell + per-window tier-fired bitmasks to compute a motif-conditional consensus boost. A mask of 0 falls back to the reason-code-derived default in affinity_tiers_for(...), preserving Phase-2 behaviour for entries without a curated mask.

confuser_motif: Option<MotifClass>

Phase 5.6 — confuser-pair adjudication. Names the primary motif expected to compete with this one for the same residual signature (e.g., DeploymentRegressionSlew’s confuser is CircuitBreakerOpenShift — both are step-shaped single-service motifs). The bank’s match function explicitly tracks the confuser’s score during scoring, reports margin_vs_confuser, and fusion gates typing on whether the candidate beats its declared confuser by margin_vs_confuser_threshold. None means no confuser is declared — episode types-confirmed based on runner-up margin alone (legacy semantics).

margin_vs_confuser_threshold: f64

Phase 5.6 — minimum margin against the declared confuser for typed confirmation. Episodes that beat the runner-up but not the confuser are reported as confuser_ambiguous rather than typed. Default 0.10 (10% of top score). Set to 0.0 to disable.

primary_witness_tiers: u32

Phase 7 — primary witness tier gate (strict semantic anti-hallucination). Subset of affinity_tiers that MUST fire for the motif to be a valid typing candidate. Distinct from the Phase 5.6 zero-tier filter (any affinity tier suffices); this requires SPECIFIC tiers to fire. E.g., DeploymentRegressionSlew witnesses = {A, B, N, X} — without scalar/Page-Hinkley/offline- CPD/Pettitt step detection actually firing, the bank refuses to type “deployment regression” even if other affinity tiers (correlation, dispersion) fired. A mask of 0 disables the gate (default — match behaves identically to Phase 5.6).

primary_witness_detectors: &'static [&'static str]

Phase 8 — per-detector primary witnesses (strict ensemble-methods SOTA gate). Named-detector subset that MUST fire for the motif to type-confirm. Distinct from primary_witness_tiers (any detector in the named tiers suffices); this requires SPECIFIC detectors (e.g., [poisson_burst, burst_after_silence] for AuthenticationFailureSpike). Names match the detector_name field returned by detector functions in incumbent_baselines.rs. Empty slice &[] disables the gate (default — Phase 7 behaviour). Applied at fusion.rs typed-confirmation, not at bank match time; failing motifs are demoted to ambiguous rather than skipped, so the runner-up motif can still be reported in the operator packet.

Trait Implementations

impl Clone for HeuristicEntry

fn clone(&self) -> HeuristicEntry

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for HeuristicEntry

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

Formats the value using the given formatter.

impl PartialEq for HeuristicEntry

fn eq(&self, other: &HeuristicEntry) -> 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 Copy for HeuristicEntry

impl StructuralPartialEq for HeuristicEntry