fix(llm): token usage mismatch during early stream aborts on Azure OpenAI models
When an Azure OpenAI stream is aborted extremely early (e.g., due to user cancellation or first-byte timeout), the usage parser occasionally assumes 0 tokens instead of aggregating the partial cache-hit tokens that were successfully transmitted in the very first delta, causing drift in telemetry metrics.
fix(llm): token usage mismatch during early stream aborts on Azure OpenAI models
When an Azure OpenAI stream is aborted extremely early (e.g., due to user cancellation or first-byte timeout), the usage parser occasionally assumes 0 tokens instead of aggregating the partial cache-hit tokens that were successfully transmitted in the very first delta, causing drift in telemetry metrics.