Struct InferenceBudget 

pub struct InferenceBudget {
    pub extract_latency_budget_ms: u32,
    pub max_inference_ms_per_commit: u32,
    pub max_phrases_embedded: u32,
    pub max_types: u32,
    pub author_rate_limit_per_commit: u32,
}

Shared primitive: wall-clock and volume caps for the inference stage of a single commit.

All fields are floor-c tunables (solution.md R6 §Constant classification table) or corpus-derived. No magic numbers.

Fields

extract_latency_budget_ms: u32

Rolling wall-clock budget for the extract-and-embed phase of inference, derived from 3 * rolling.p50_ingest_phrase_embed_ms (fallback 500ms). Class a (rolling telemetry).

max_inference_ms_per_commit: u32

Hard wall-clock ceiling per commit. Class c (reference standard): 250ms = 50% of max_cooccurrence_ms = 500.

max_phrases_embedded: u32

Hard ceiling on bridging-phrase embeddings per commit. Class a (corpus-derived): min(50_000, sqrt(N_phrases) * 100).

max_types: u32

Max inferred relation types emitted per commit. Class a (corpus-derived): ceil(log2(corpus_size)).

author_rate_limit_per_commit: u32

Per-author bridging-phrase cap per commit. Class a (corpus-derived): max(200, 0.01 * mean_phrases_per_author).

Implementations

impl InferenceBudget

pub const MAX_INFERENCE_MS_PER_COMMIT: u32 = 250

Floor-c reference standard for the per-commit hard wall.

250ms. See solution.md R6 §Floor-c apparatus.

pub const FALLBACK_EXTRACT_LATENCY_MS: u32 = 500

Fallback for extract_latency_budget_ms when rolling p50 telemetry is unavailable. See shared/inference-budget.md §API sketch.

pub const fn conservative() -> Self

Conservative defaults for CI / proptest / initial runs.

Real deployments derive via a stats-aware constructor in the ingest crate; this keeps mnem-extract free of clock or telemetry dependencies.

pub const fn effective_ms(&self) -> u32

Effective runtime budget = min(extract_latency_budget_ms, max_inference_ms_per_commit).

The hard wall is always the ceiling, even when rolling telemetry computes a higher extract budget.

pub fn effective_ms_gauge(&self) -> (&'static str, f64)

Sample for gauge mnem_inference_budget_effective_ms.

Caller is responsible for emission; this module does not link a metrics backend. Returning (name, value) lets the caller use either metrics::gauge! or a custom registry.

pub fn validate(&self) -> Result<(), &'static str>

Validate that the budget is internally consistent.

Returns Err with a static reason when:

1. max_inference_ms_per_commit is zero.
2. extract_latency_budget_ms is zero.
3. max_phrases_embedded is zero.

Zero-valued caps are always a programming error: they would make the entire inference pass a no-op and silently hide bugs.

Trait Implementations

impl Clone for InferenceBudget

fn clone(&self) -> InferenceBudget

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for InferenceBudget

impl Debug for InferenceBudget

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

Formats the value using the given formatter.

impl Default for InferenceBudget

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for InferenceBudget

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for InferenceBudget

impl PartialEq for InferenceBudget

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for InferenceBudget

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 InferenceBudget