Module cost

Structs§

ToolLoopCostTrackingContext: Context for cost tracking within the tool call loop. Scoped via tokio::task_local! at call sites (channels, gateway).
TurnUsage: Per-scope token/cost accumulator derived from the usage events emitted during a single task-local runtime invocation.
UnpricedModel: A model with token usage whose ledger records explicitly report that one or more token-bearing pricing dimensions were unavailable.

build_model_provider_pricing
build_type_level_model_provider_pricing
check_tool_loop_budget: Check budget before an LLM call. Returns None when no cost tracking context is scoped (tests, delegate, CLI without cost config).
compute_cost_usd: Compute cost_usd for a provider/model/usage triple using the task-local pricing context but WITHOUT updating any accumulators. For emitting Usage events for rejected attempts where the cost is already accounted via settle_provider_attempts. Prices the usage exactly like the settlement path, including the cache-write band, so the projected rejected cost and the settled record agree.
provider_pricing
record_rejected_tool_loop_cost_usage: Record usage from a rejected provider attempt without replacing the accepted response’s context-window fill.
record_tool_loop_cost_usage: Record token usage from an accepted LLM response via the task-local cost tracker. Returns (total_tokens, cost_usd) on success, None when not scoped or no usage.
tool_loop_cost_tracking_context_for_agent
tool_loop_cost_tracking_context_from_tracker
unpriced_models_in_summary: Return per-model unpriced token subsets from the ledger-derived summary, sorted by model id for stable output.