agent how to count the latency of each flow

[joeylin]

The latency of each process is measured, for example, from the time the user speaks to the time when the content is recognized by the ASR, to the time when the first text is given to the LLM, and the first token is output by the llm