Struct AdmissionConfig 

pub struct AdmissionConfig {
    pub enabled: bool,
    pub threshold: f32,
    pub fast_path_margin: f32,
    pub admission_provider: ProviderName,
    pub weights: AdmissionWeights,
    pub admission_strategy: AdmissionStrategy,
    pub rl_min_samples: u32,
    pub rl_retrain_interval_secs: u64,
    pub goal_conditioned_write: bool,
    pub goal_utility_provider: ProviderName,
    pub goal_utility_threshold: f32,
    pub goal_utility_weight: f32,
}

Configuration for A-MAC adaptive memory admission control ([memory.admission] TOML section).

When enabled = true, a write-time gate evaluates each message before saving to memory. Messages below the composite admission threshold are rejected and not persisted.

Fields

enabled: bool

Enable A-MAC admission control. Default: false.

threshold: f32

Composite score threshold below which messages are rejected. Range: [0.0, 1.0]. Default: 0.40.

fast_path_margin: f32

Margin above threshold at which the fast path admits without an LLM call. Range: [0.0, 1.0]. When heuristic score >= threshold + margin, LLM call is skipped. Default: 0.15.

admission_provider: ProviderName

Provider name from [[llm.providers]] for future_utility LLM evaluation. Falls back to the primary provider when empty. Default: "".

weights: AdmissionWeights

Per-factor weights. Normalized at runtime. Default: {0.30, 0.15, 0.30, 0.10, 0.15}.

admission_strategy: AdmissionStrategy

Admission decision strategy. Default: heuristic.

rl_min_samples: u32

Minimum training samples before the RL model is activated. Below this count the system falls back to Heuristic. Default: 500.

rl_retrain_interval_secs: u64

Background RL model retraining interval in seconds. Default: 3600.

goal_conditioned_write: bool

Enable goal-conditioned write gate (#2408). When true, memories are scored against the current task goal and rejected if relevance is below goal_utility_threshold. Zero regression when false. Default: false.

goal_utility_provider: ProviderName

Provider name from [[llm.providers]] for goal-utility LLM refinement. Used only for borderline cases (similarity within 0.1 of threshold). Falls back to the primary provider when empty. Default: "".

goal_utility_threshold: f32

Minimum cosine similarity between goal embedding and candidate memory to consider it goal-relevant. Below this, goal_utility = 0.0. Default: 0.4.

goal_utility_weight: f32

Weight of the goal_utility factor in the composite admission score. Set to 0.0 to disable (equivalent to goal_conditioned_write = false). Default: 0.25.

Trait Implementations

impl Clone for AdmissionConfig

fn clone(&self) -> AdmissionConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for AdmissionConfig

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

Formats the value using the given formatter.

impl Default for AdmissionConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for AdmissionConfig

where AdmissionConfig: Default,

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for AdmissionConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.