Resilience

The public resilience facade is victor.framework.resilience. It re-exports provider
circuit breakers and the shared retry strategies; the removed
victor.observability.resilience module is not an alternative import path.

Retry strategies

from victor.framework.resilience import ExponentialBackoffStrategy, with_retry

@with_retry(ExponentialBackoffStrategy(max_attempts=3))
async def fetch_with_retry():
    # Replace with your asynchronous operation.
    return "ready"

Other exported strategies include LinearBackoffStrategy, FixedDelayStrategy, and
NoRetryStrategy. RetryExecutor supports explicit execution instead of a decorator.
Choose retryable failures and limits appropriate to the operation: retrying an external
write may repeat its side effects.

Provider resilience

ResilientProvider combines a primary provider, optional fallback providers, circuit
configuration, and provider retry configuration. Configure it when constructing runtime
provider services. These are provider objects, not arbitrary Agent.create keywords.
The facade exports CircuitBreaker, CircuitBreakerRegistry, CircuitBreakerConfig,
and ProviderRetryConfig for that integration.

Observability

Use topic-based events for runtime instrumentation. A
telemetry bus is not a bulkhead, rate limiter, or durable retry queue. The earlier guide
advertised unimplemented Bulkhead, RateLimiter, and category-based bus helpers;
those examples are preserved only in page history.

Circuit-breaker lifecycle

The circuit-breaker recovery cycle moves between closed, open, and probe states.

---
title: Circuit-breaker recovery cycle
---
%%{init: {"theme":"base","themeVariables":{"primaryColor":"#E8EFF7","primaryTextColor":"#17324D","primaryBorderColor":"#456987","lineColor":"#456987","fontFamily":"Arial"}}}%%
stateDiagram-v2
  [*] --> Closed
  Closed --> Open: failure threshold reached
  Open --> HalfOpen: recovery timeout elapsed
  HalfOpen --> Closed: success threshold reached
  HalfOpen --> Open: probe failure