The latency is much smaller than the actual execution time #190
Description
The latency is much smaller than the actual execution time because each LLM call doesn't have its own started_at timestamp— they all share the same class variable.
{{ message }}