Struct TokenTracker 

pub struct TokenTracker { /* private fields */ }

Tracks token usage from LLM API responses. Uses real usage data from API, not estimation.

Cumulative totals are stored for the dimensions consumers care about today (input/output, cache read/creation, reasoning). The last_usage field preserves the full TokenUsage from the most recent API call, so audio / video / prediction dimensions are still accessible without growing dedicated accumulators until a consumer asks for them.

Implementations

impl TokenTracker

pub fn new(context_limit: Option<u32>) -> Self

pub fn record_usage(&mut self, sample: TokenUsage)

Record usage from an LLM API response.

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

Current context usage as a ratio (0.0 - 1.0)

pub fn exceeds_threshold(&self, threshold: f64) -> bool

Whether current usage exceeds the given threshold

pub fn should_compact(&self, threshold: f64) -> bool

Check if context should be compacted based on the given threshold. This is a convenience method that combines usage ratio check with a minimum context size requirement to avoid unnecessary compaction on small conversations.

pub fn tokens_remaining(&self) -> Option<u32>

Tokens remaining before hitting limit

pub fn set_context_limit(&mut self, limit: Option<u32>)

Update the context limit (e.g. when switching models)

pub fn context_limit(&self) -> Option<u32>

Get the context limit

pub fn last_input_tokens(&self) -> u32

Get last recorded input tokens (current context size)

pub fn last_usage(&self) -> &TokenUsage

Get the full TokenUsage from the most recent API call. Returns the default (all zeros / None) before any call has been recorded.

pub fn total_input_tokens(&self) -> u64

Get total input tokens across all calls

pub fn total_output_tokens(&self) -> u64

Get total output tokens across all calls

pub fn total_cache_read_tokens(&self) -> u64

Get total cache-read tokens across all calls

pub fn total_cache_creation_tokens(&self) -> u64

Get total cache-creation tokens across all calls

pub fn total_reasoning_tokens(&self) -> u64

Get total reasoning tokens across all calls

pub fn reset_current_usage(&mut self)

Reset current usage tracking after context compaction. Preserves cumulative totals for metrics while clearing last_usage to prevent immediate re-triggering of compaction.

Trait Implementations

impl Clone for TokenTracker

fn clone(&self) -> TokenTracker

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TokenTracker

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

Formats the value using the given formatter.

impl Default for TokenTracker

fn default() -> TokenTracker

Returns the “default value” for a type.