Retry and fallback logic in Agentic AI

Introduction

Retry and fallback logic helps your AI keep trying when something goes wrong and use a backup plan if needed. This makes your AI more reliable and user-friendly.

When your AI calls an external service that might be slow or fail sometimes.

When your AI tries to get data but the first method might not always work.

When you want to avoid stopping your AI because of small errors.

When you want to improve user experience by handling errors smoothly.

Syntax

Agentic AI

def retry_and_fallback(task, retries=3, fallback=None):
    for attempt in range(retries):
        try:
            return task()
        except Exception as e:
            print(f"Attempt {attempt + 1} failed: {e}")
    if fallback:
        return fallback()
    raise Exception("All retries failed and no fallback available.")

The task is the main function you want to try.

The fallback is a backup function if all retries fail.

Examples

This tries the task twice, then uses fallback if both fail.

Agentic AI

def task():
    # Try to get data
    pass

def fallback():
    # Return default data
    pass

result = retry_and_fallback(task, retries=2, fallback=fallback)

This retries a task that always fails (division by zero) three times, then returns 'default'.

Agentic AI

result = retry_and_fallback(lambda: 1 / 0, retries=3, fallback=lambda: 'default')

Sample Model

This program tries to run unreliable_task up to 5 times. If it keeps failing, it uses fallback_task. It prints each failure and the final result.

Agentic AI

def unreliable_task():
    import random
    if random.random() < 0.7:
        raise ValueError("Random failure")
    return "Success"

def fallback_task():
    return "Fallback result"

def retry_and_fallback(task, retries=3, fallback=None):
    for attempt in range(retries):
        try:
            return task()
        except Exception as e:
            print(f"Attempt {attempt + 1} failed: {e}")
    if fallback:
        return fallback()
    raise Exception("All retries failed and no fallback available.")

result = retry_and_fallback(unreliable_task, retries=5, fallback=fallback_task)
print(f"Final result: {result}")

Important Notes

Set retries to a reasonable number to avoid long waits.

Fallback should be simple and safe to run.

Print or log errors to understand failures during retries.

Summary

Retry logic tries a task multiple times to handle temporary problems.

Fallback logic provides a backup plan if retries fail.

Together, they make AI systems more reliable and user-friendly.