Fallback Tiers

Fallback tiers are your AI providers. They form an ordered failover chain, and the first enabled tier is what chat streams from.

The chain

Open the Fallback view (shield icon). Each tier shows:

• Provider + model
• Base URL (OpenAI-compatible endpoint)
• API key (stored in the encrypted vault)
• Timeout and per-tier enable toggle

Reorder tiers with the arrows, add tiers with Add Fallback Tier, and edit the selected tier on the right. Save Chain persists the configuration.

How routing uses the chain

• Chat: the first enabled tier receives the request directly — model, endpoint, and API key are all passed through, so streaming authenticates automatically.
• DAG tasks: each role uses the tier you assigned to it (see DAG Pipeline); Inherit from Auto Router defers to the router.
• If a request fails, the chain concept applies: later enabled tiers are the fallback targets.

Chain presets

Save the current chain under a name (e.g. Cheap nightly, Quality first):

• Save chain as preset — stores every tier with its settings.
• Apply to active chain — swaps the live chain to a saved preset.
• Per execution mode, assign a different preset and optionally pair it with Auto-Continue — activating the mode then also switches the chain.

Import / export

Chains round-trip as JSON: Import Config / Export Config in the header. This makes it easy to share a provider setup between machines.