Struct TokenUsageRecord 

pub struct TokenUsageRecord {

    pub ts: String,
    pub session_id: String,
    pub model: String,
    pub provider: String,
    pub message_count: usize,
    pub cache_creation_input_tokens: u64,
    pub cache_read_input_tokens: u64,
    pub input_tokens: u64,
    pub output_tokens: u64,
    pub thinking_tokens: u64,
    pub system_tokens: u32,
    pub summary_tokens: u32,
    pub window_tokens: u32,
    pub total_tokens: u32,
    pub max_context_tokens: u32,
    pub budget_limit: u32,
    pub prompt_cached_tool_outputs: usize,
    pub prompt_cached_tool_tokens_saved: u32,
    pub truncation_occurred: bool,
    pub segments_removed: usize,
}

One per-LLM-call usage record. Flattened for line-oriented analysis (jq / DuckDB / pandas over the JSONL).

Fields

ts: String

RFC3339 wall-clock timestamp captured when the call completed.

session_id: String

model: String

provider: String

message_count: usize

Conversation length (message count) at emit time — a monotonic-ish ordinal to correlate records with conversation growth.

cache_creation_input_tokens: u64

cache_read_input_tokens: u64

input_tokens: u64

Non-cached “fresh” input tokens (server-reported), disjoint from the two cache counts. The precise prompt size is input_tokens + cache_read + cache_creation, and the exact cache-hit ratio is cache_read / that_sum.

output_tokens: u64

thinking_tokens: u64

system_tokens: u32

summary_tokens: u32

window_tokens: u32

total_tokens: u32

max_context_tokens: u32

budget_limit: u32

prompt_cached_tool_outputs: usize

prompt_cached_tool_tokens_saved: u32

truncation_occurred: bool

segments_removed: usize

Implementations

impl TokenUsageRecord

pub fn new( ts: String, session_id: &str, model: &str, provider: &str, message_count: usize, usage: Option<&TokenBudgetUsage>, cache_creation_input_tokens: u64, cache_read_input_tokens: u64, input_tokens: u64, output_tokens: u64, thinking_tokens: u64, ) -> Self

Build a record from the prompt-side budget snapshot (usage) and the server-side stream stats. The cache-creation count lives only on the stream output — it is not part of TokenBudgetUsage — so it is passed in explicitly.

pub fn to_json_line(&self) -> Result<String, Error>

Serialize to a single-line JSON string (no trailing newline; the storage layer frames the line).

Trait Implementations

impl Clone for TokenUsageRecord

fn clone(&self) -> TokenUsageRecord

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for TokenUsageRecord

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

Formats the value using the given formatter.

impl Serialize for TokenUsageRecord

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

where __S: Serializer,

Serialize this value into the given Serde serializer.