Type Alias ContextTurn

Per-turn token snapshot — used to render context window fill over time. Each entry represents one deduplicated API call.

type ContextTurn = {
    cacheCreation: number;
    cacheRead: number;
    input: number;
    output: number;
    ts: string;
}

Properties

cacheCreation: number

cache_creation_input_tokens — prompt tokens written to cache.

cacheRead: number

cache_read_input_tokens — prompt tokens served from cache.

input: number

Raw input_tokens for this turn (excludes cache).

output: number

output_tokens generated by the model.

ts: string

ISO timestamp of the assistant turn.