Struct RoutingConfig 

pub struct RoutingConfig {

    pub mode: String,
    pub confidence_threshold: f64,
    pub local_first: bool,
    pub cost_aware: bool,
    pub estimated_output_tokens: u32,
    pub accuracy_floor: f64,
    pub accuracy_min_obs: usize,
    pub cost_weight: Option<f64>,
    pub canary_model: Option<String>,
    pub canary_fraction: f64,
    pub blocked_models: Vec<String>,
    pub per_provider_timeout_seconds: u64,
    pub max_total_inference_seconds: u64,
    pub max_fallback_attempts: usize,
}

Fields

mode: String

confidence_threshold: f64

local_first: bool

cost_aware: bool

estimated_output_tokens: u32

accuracy_floor: f64

Minimum observed quality score (0.0–1.0) for a model to be considered during metascore routing. Models with fewer than accuracy_min_obs observations are exempt (insufficient data). Set to 0.0 to disable.

accuracy_min_obs: usize

Minimum observations before the accuracy floor applies to a model.

cost_weight: Option<f64>

Custom cost weight for metascore [0.0–1.0]. When set, replaces the binary cost_aware toggle with a continuous dial: 0.0 = ignore cost, 1.0 = maximize savings. Efficacy weight adjusts inversely. When None, falls back to cost_aware boolean behavior.

canary_model: Option<String>

Canary model to route a fraction of traffic through for A/B validation. When set, canary_fraction of requests are routed to this model instead of the metascore winner. Set to None to disable canary routing.

canary_fraction: f64

Fraction of requests routed to the canary model [0.0–1.0]. Only effective when canary_model is set. Default: 0.0 (disabled).

blocked_models: Vec<String>

Static model blocklist — models listed here are unconditionally excluded from all routing paths (override, metascore, fallback). Useful as an instant kill-switch without restarting the server.

per_provider_timeout_seconds: u64

Per-provider timeout in seconds for interactive inference. If a single model doesn’t respond within this window, the fallback chain advances. Increase for slow local models (e.g. large quantized models on CPU/GPU).

max_total_inference_seconds: u64

Total wall-clock budget in seconds for the entire inference fallback chain (all attempts combined). Increase if you have many fallback candidates or slow providers.

max_fallback_attempts: usize

Maximum number of fallback attempts before giving up.

Trait Implementations

impl Clone for RoutingConfig

fn clone(&self) -> RoutingConfig

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RoutingConfig

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

Formats the value using the given formatter.

impl Default for RoutingConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for RoutingConfig

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for RoutingConfig

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

where __S: Serializer,

Serialize this value into the given Serde serializer.