Struct Usage 

pub struct Usage {
    pub prompt_tokens: u32,
    pub completion_tokens: u32,
    pub total_tokens: u32,
    pub cached_tokens: Option<u32>,
}

Token usage for one conversation.

Field names match the OpenAI wire format; total_tokens follows the vendor’s convention (not necessarily the sum of the other two). Default = all zeros, with nothing reported.

Presence is carried by the enclosing type: ChatResponse::usage / StreamEvent::Done::usage are Option<Usage> — None means the endpoint did not report usage for this turn, Some means the reported values. The distinction matters for observability: “not reported” is not the same as “reportedly zero” (the Agent layer also tracks it in RunSummary::usage_omitted).

cached_tokens carries the same distinction inside a reported usage: the prompt-cache breakdown is optional per turn even when the totals are present.

Fields

prompt_tokens: u32

Input tokens for this turn.

completion_tokens: u32

Output tokens for this turn.

total_tokens: u32

Total for this turn (vendor convention).

cached_tokens: Option<u32>

Prompt tokens served from the provider’s prompt cache.

A subset of prompt_tokens, not an addition to it (vendors bill cached input at a lower rate).

None = the endpoint did not report a cache breakdown for this turn; Some(0) = it reported that nothing hit the cache. Endpoints differ here — prompt-caching ones report the breakdown (OpenAI / DeepSeek as prompt_tokens_details.cached_tokens), others omit it — so a reported zero must not be inferred from an omission. When turns are summed, only the reported parts contribute and RunSummary::cache_omitted marks whether the sum is a lower bound.

Implementations

impl Usage

pub fn new(prompt_tokens: u32, completion_tokens: u32) -> Self

Constructs from input / output counts; the total is summed automatically and no cache breakdown is reported (with_cached_tokens sets one).

pub fn with_cached_tokens(self, cached_tokens: u32) -> Self

Sets the prompt tokens served from the provider’s cache, marking the breakdown as reported (Some in cached_tokens).

Trait Implementations

impl AddAssign for Usage

Usage accumulates per turn (the Agent layer sums tokens across turns).

cached_tokens sums the parts that were reported: a turn that omitted the breakdown contributes nothing rather than zero.

fn add_assign(&mut self, rhs: Self)

Performs the += operation.

impl Clone for Usage

fn clone(&self) -> Self

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Copy for Usage

impl Debug for Usage

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

Formats the value using the given formatter.

impl Default for Usage

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for Usage

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for Usage

impl PartialEq for Usage

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

Equality operator ==.

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

Inequality operator !=.

impl Serialize for Usage

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 Usage