Struct CacheShareConfig 

pub struct CacheShareConfig {
    pub alarm_threshold: Option<f64>,
    pub window_secs: Option<u64>,
    pub min_prompt_tokens: Option<u64>,
    pub open_model_providers: Vec<String>,
    pub open_model_patterns: Vec<String>,
    pub open_model_policy: Option<String>,
}

[proxy.cache_share] — guards the provider prompt-cache discount.

Without prompt caching a long agent session bills every prompt token at the full input rate; with it most prompt tokens bill at the cache-read rate. The proxy keeps a rolling window per provider+model and raises an alarm when the cache-read share of the prompt tokens drops below alarm_threshold.

Open-weight models (Llama, Qwen, DeepSeek, …) are served by many hosts, and not all of them give a reliable cache discount. A provider not in open_model_providers is flagged (open_model_policy = "flag", the default) or refused before the request leaves ("refuse"). Local endpoints are exempt: nothing is bought there.

Fields

alarm_threshold: Option<f64>

Minimum cache-read share (0–1) of the prompt tokens; default 0.90.

window_secs: Option<u64>

Rolling window in seconds; default 3600, capped at 7 days.

min_prompt_tokens: Option<u64>

Prompt tokens a window needs before the alarm may fire, so the cold first turn of a session (all cache writes) does not alarm; default 200 000.

open_model_providers: Vec<String>

Provider ids (registry id or built-in label, case-insensitive) with a reliable cache discount, allowed to serve open models.

open_model_patterns: Vec<String>

Extra model-name fragments (case-insensitive) counted as open models, added to the built-in list.

open_model_policy: Option<String>

"flag" (default) or "refuse" for an open model on a provider outside open_model_providers.

Implementations

impl CacheShareConfig

pub fn effective_threshold(&self) -> f64

Configured threshold clamped to 0–1, or the 0.90 default.

pub fn effective_window_secs(&self) -> u64

Configured window clamped to 60 s – 7 days, or one hour.

pub fn effective_min_prompt_tokens(&self) -> u64

Configured minimum prompt tokens, or the default.

pub fn refuses_unlisted_open_models(&self) -> bool

True when an unlisted provider must refuse an open model; any value other than "refuse" flags only.

Trait Implementations

impl Clone for CacheShareConfig

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for CacheShareConfig

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

Formats the value using the given formatter.

impl Default for CacheShareConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for CacheShareConfig

where CacheShareConfig: Default,

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl PartialEq for CacheShareConfig

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for CacheShareConfig

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 CacheShareConfig