hook-model-fallback

Model fallback automates switching to a pre-configured backup model when the primary model fails due to rate limits or outages, ensuring continuous chat workflow availability without manual intervention.

The best way to configure automatic model switching is by integrating a hook that initializes with your chosen fallback models and retry settings, tracking failure counts to automatically trigger backups on demand.

Yes, you can track model failure counts per session. This mechanism maintains consistency by automatically resetting the failure tracking and retry limits whenever a new chat session begins.

Per-session failure tracking works by monitoring primary model errors within an active chat and resetting the failure counts when a new session starts, preventing permanent fallback states across unrelated conversations.

No, you do not need to manually switch models during intermittent API outages. The hook automatically detects primary model failures and transitions to configured alternative models based on your retry limits.

You should not use automatic model fallback if your application requires strict consistency from a single model, as it will automatically switch to a different backup model during rate limits or degraded responses.