A circuit breaker temporarily stops routing to a provider that is failing repeatedly, rather than letting every request pay the latency and cost of a doomed call. Combined with fallback chains, circuit breakers keep latency and error rates stable when a provider degrades.
Why it matters
Retrying a provider that is already failing wastes latency and budget. A breaker stops the bleeding by skipping a provider that is down.
How it works
After repeated failures the breaker "opens" and traffic bypasses that provider for a cooldown, then retries a probe request to see if it recovered. Combined with fallbacks, it keeps error rates and latency stable during an outage.