model-routing

Reroute AI models during a vendor outage by applying time-boxed routing overrides using a tiered fallback matrix. This ensures service continuity by temporarily redirecting traffic from affected models to safe fallback targets.

Manage fallback routing for tiered models by consulting a same-tier fallback matrix to select appropriate targets. This protocol coordinates overrides specifically for tiers like Opus, Sonnet, and Haiku to maintain service during disruptions.

ADR records work with model routing overrides by enforcing a structured workflow that opens and closes an Architecture Decision Record for every override. This enforces time caps and requires renewal ADRs for expiry extensions.

Trace session logs for agents affected by a model outage by automatically tagging the session logs with specific override IDs. This tagging guarantees full traceability for all interactions impacted by the temporary routing changes.

You can enforce time limits on temporary model fallback overrides through time-boxed caps. The system enforces strict expiry durations and requires a renewal ADR if the vendor outage continues and the override needs extension.

Use a structured protocol for model routing changes when a vendor outage is detected and you need to specify the affected tier. It ensures safe reversion to the original model once the vendor service fully recovers.