Latency monitoring per step in Agentic Ai

Introduction

Latency monitoring per step helps you see how long each part of a process takes. This helps find slow parts and fix them.

You want to improve the speed of a multi-step task.

You need to find which step causes delays in your system.

You want to track performance changes after updates.

You want to report detailed timing to your team or users.

Syntax

Agentic_ai

monitor_latency(step_name: str) -> float

Call this function at the start or end of each step.

It returns the time taken in seconds for that step.

Examples

Measure how long 'step1' takes by recording start and end times.

Agentic_ai

start_time = monitor_latency('step1')
# do step 1 work
end_time = monitor_latency('step1')
latency = end_time - start_time

Get latency for a database query step and print it.

Agentic_ai

latency = monitor_latency('database_query')
print(f"Database query took {latency:.2f} seconds")

Sample Program

This program measures how long two steps take by recording start and end times using the monitor_latency function.

Agentic_ai

import time

def monitor_latency(step_name: str, start_times={}):
    now = time.time()
    if step_name in start_times:
        latency = now - start_times.pop(step_name)
        return latency
    else:
        start_times[step_name] = now
        return 0.0

# Start step 1
monitor_latency('step1')
time.sleep(1.2)  # simulate work
# End step 1
latency1 = monitor_latency('step1')

# Start step 2
monitor_latency('step2')
time.sleep(0.8)  # simulate work
# End step 2
latency2 = monitor_latency('step2')

print(f"Step 1 latency: {latency1:.2f} seconds")
print(f"Step 2 latency: {latency2:.2f} seconds")

Important Notes

Always call monitor_latency first to start timing, then call it again to get the latency.

Use unique step names to avoid confusion.

Latency is measured in seconds with decimals for precision.

Summary

Latency monitoring per step helps find slow parts in a process.

Call the function at start and end of each step to measure time.

Use the results to improve system speed and user experience.