Skip to main content

CODITECT Retry Engine - Resilient Operation Execution with Backoff

Part of Track H.2.5: Inter-Agent Communication Infrastructure Based on AUTONOMOUS-AGENT-SYSTEM-DESIGN.md specifications

This module provides:

  • RetryEngine: Configurable retry execution with multiple backoff strategies
  • RetryPolicy: Delay calculation with jitter support
  • @retry decorator: Easy function wrapping
  • RetryWithCircuitBreaker: Coordinated fault tolerance
  • Comprehensive metrics and monitoring

Backoff Strategies: ┌─────────────────────────────────────────────────────────────────┐ │ EXPONENTIAL: delay = base * (multiplier ^ attempt) │ │ LINEAR: delay = base + (increment * attempt) │ │ FIXED: delay = base (constant) │ │ FIBONACCI: delay = base * fib(attempt) │ │ DECORRELATED: delay = random(base, previous_delay * 3) │ └─────────────────────────────────────────────────────────────────┘

Jitter Types (prevent thundering herd): ┌─────────────────────────────────────────────────────────────────┐ │ NONE: No jitter (exact delay) │ │ FULL: random(0, delay) │ │ EQUAL: delay/2 + random(0, delay/2) │ │ DECORRELATED: random(base, delay * 3) - AWS recommended │ └─────────────────────────────────────────────────────────────────┘

Usage: from scripts.core.retry_engine import ( RetryEngine, RetryConfig, retry, RetryWithCircuitBreaker, )

# Using decorator
@retry(max_attempts=5, backoff="exponential")
async def call_api(endpoint: str):
    return await http_client.get(endpoint)

# Using engine directly
engine = RetryEngine(RetryConfig(
    max_attempts=5,
    base_delay=1.0,
    max_delay=60.0,
    backoff_strategy=BackoffStrategy.EXPONENTIAL,
    jitter=JitterType.FULL,
))

result = await engine.execute(risky_function, arg1, arg2)

# With circuit breaker integration
resilient = RetryWithCircuitBreaker(
    retry_config=RetryConfig(max_attempts=3),
    circuit_breaker=breaker,
)
result = await resilient.execute(api_call)

Author: CODITECT Framework Created: January 8, 2026 Version: 1.0.0

File: retry_engine.py

Classes​

BackoffStrategy​

Backoff strategy for retry delays.

JitterType​

Jitter type to add randomness to delays.

RetryDecision​

Decision for whether to retry.

RetryError​

Base exception for retry errors.

MaxRetriesExceeded​

Raised when max retries exceeded.

RetryTimeout​

Raised when total retry time exceeded.

RetryConfig​

Configuration for retry behavior.

RetryAttempt​

Record of a single retry attempt.

RetryMetrics​

Metrics for retry operations.

RetryPolicy​

Calculates retry delays based on configuration.

Functions​

retry(max_attempts, base_delay, max_delay, multiplier, backoff, jitter, retry_exceptions, ignore_exceptions, total_timeout, on_retry)​

Decorator to add retry logic to a function.

create_retry_engine(max_attempts, base_delay, backoff, jitter)​

Create a retry engine with common settings.

exponential_backoff(max_attempts, base_delay, max_delay, multiplier)​

Create retry engine with exponential backoff.

linear_backoff(max_attempts, base_delay, max_delay, increment)​

Create retry engine with linear backoff.

constant_backoff(max_attempts, delay)​

Create retry engine with constant delay.

retry_sync(func)​

Convenience function to retry a sync function.

get_retry_config(name)​

Get a predefined retry configuration.

to_dict()​

Convert to dictionary (without callables).

calculate_delay(attempt)​

Calculate delay for a given attempt number.

should_retry(attempt, exception, elapsed)​

Decide whether to retry.

reset()​

Reset policy state (for decorrelated backoff).

execute_sync(func)​

Execute a synchronous function with retry logic.

get_metrics()​

Get retry metrics.

reset_metrics()​

Reset all metrics.

execute_sync(func)​

Execute synchronously with retry and circuit breaker.

Usage​

python retry_engine.py