hook-runtime-fallback

Switch to available fallback providers during long-running conversations by configuring an auto-retry hook with a prioritized fallback model list. This rotates across multi-provider setups to maintain session availability when the primary provider temporarily fails.

Runtime fallback reacts to transient errors like HTTP 429 rate limits during an active session, whereas session start model fallback handles model unavailability before the conversation begins. Runtime fallback auto-switches models in real-time to keep ongoing sessions running.

Yes, you can configure cooldown periods for AI model fallback rotation. The runtime fallback hook exposes configuration options for cooldowns and rotation modes, allowing you to control exactly how fallbacks rotate and recover after transient runtime errors occur.

Runtime fallback works across multi-provider AI setups to improve reliability. It automatically detects runtime errors and selects alternative models from a prioritized fallback list, ensuring continuous availability across different providers during long sessions.

Automatic fallback to another AI model is triggered by transient runtime errors such as HTTP 429 rate limits, overload, timeout, and capacity errors. The hook detects these errors in real-time and rotates to the next available provider.