Struct TokenUsage 

pub struct TokenUsage {
    pub input_tokens: u32,
    pub output_tokens: u32,
    pub cache_read_tokens: Option<u32>,
    pub cache_creation_tokens: Option<u32>,
    pub actual_cost_usd: Option<f64>,
    pub estimated_cost_usd: Option<f64>,
    pub effective_cost_usd: Option<f64>,
}

Token usage statistics

Tracks token consumption per LLM call including cache tokens for cost optimization.

Disjoint bucket convention

Prompt token buckets are disjoint (non-overlapping). Drivers normalize provider wire formats at the boundary so this holds for every provider:

total_prompt = input_tokens + cache_read_tokens + cache_creation_tokens

• input_tokens — non-cached prompt tokens only.
• cache_read_tokens — tokens served from cache, never counted in input_tokens.
• cache_creation_tokens — tokens written to cache, never counted in input_tokens.

Inclusive providers (OpenAI Responses / Chat Completions, Gemini) report a prompt count that includes cached reads; their drivers subtract the cached subset so the value stored here is the non-cached remainder. Anthropic / Bedrock already report disjoint buckets. Cost is therefore uniform across providers (input·in + cache_read·cr + cache_creation·cw + output·out); consumers must not re-derive a non-cached input by subtracting cache reads.

Fields

input_tokens: u32

Number of non-cached prompt tokens (cached reads/writes are tracked separately; see the disjoint bucket convention on the struct)

output_tokens: u32

Number of output/completion tokens

cache_read_tokens: Option<u32>

Number of tokens read from cache (reduces cost), disjoint from input_tokens

cache_creation_tokens: Option<u32>

Number of tokens written to cache, disjoint from input_tokens

actual_cost_usd: Option<f64>

Actual cost of this generation in USD, as reported by the provider inline (e.g. OpenRouter’s usage.cost, which reflects real post-routing/BYOK/cache pricing). None for providers that do not return a cost.

estimated_cost_usd: Option<f64>

Estimated cost of this generation in USD, derived from the model’s static price-table profile. Computed whenever a profile with cost data exists, independently of actual_cost_usd, so estimate-vs-actual drift can be reconciled. None when there is no profile cost data for the model.

effective_cost_usd: Option<f64>

Best-effort USD cost when it cannot be derived from actual/estimated alone: already-aggregated usage, or a generation carrying a cost that belongs to neither slot. Per-generation usage normally leaves this unset and derives the effective cost from actual-else-estimated; the exception is a turn whose compaction cost is folded in, where the combined total has to live here precisely so the generation’s own actual-vs-estimated distinction survives (EVE-895).

Implementations

impl TokenUsage

pub fn new(input_tokens: u32, output_tokens: u32) -> Self

Create a new TokenUsage with just input and output tokens

pub fn with_cache( input_tokens: u32, output_tokens: u32, cache_read_tokens: Option<u32>, cache_creation_tokens: Option<u32>, ) -> Self

Create a TokenUsage with cache tokens

pub fn with_cost( self, actual_cost_usd: Option<f64>, estimated_cost_usd: Option<f64>, ) -> Self

Set the actual (provider-reported) and estimated (price-table) USD costs, returning self for chaining. The two are tracked independently so an authoritative charge stays distinguishable from an estimate.

pub fn with_effective_cost(self, effective_cost_usd: Option<f64>) -> Self

Set the precomputed best-effort USD cost, returning self for chaining. Generation usage should leave this unset so the best-effort cost remains actual-else-estimated for that generation, unless a cost outside those two slots is folded in (see the field’s documentation).

pub fn effective_cost_usd(&self) -> Option<f64>

Best-effort cost in USD. Aggregate usage can carry a precomputed total that sums actual-else-estimated per generation; otherwise this falls back to the generation-level actual-else-estimated behavior.

pub fn total_tokens(&self) -> u32

Get total tokens (input + output)

pub fn add(&mut self, other: &TokenUsage)

Add another TokenUsage to this one (for aggregation)

Trait Implementations

impl Clone for TokenUsage

fn clone(&self) -> TokenUsage

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TokenUsage

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

Formats the value using the given formatter.

impl Default for TokenUsage

fn default() -> TokenUsage

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for TokenUsage

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

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for TokenUsage

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

where __S: Serializer,

Serialize this value into the given Serde serializer.