onAfterModelInvocation

The onAfterModelInvocation event fires after the AI message has been fully processed, added to the history, and token usage statistics are available.

Usage

This event provides the final snapshot of the LLM call. It is the primary location for extracting token metrics for billing or observability.

Payload Details

agent.onAfterModelInvocation(({ message, usage, finish_reason, context }) => {
    console.log("Tokens used:", usage?.total_tokens);
    console.log("Finish reason:", finish_reason);
});