omniroute-routing

AI model failover automatically switches provider requests during an outage by defining a combo with a priority or auto strategy. You specify target provider and model pairs so routing decisions adapt and redirect traffic to available targets.

To load balance requests across multiple AI models, create a named combo using strategies like weighted, round-robin, or least-used. This distributes incoming API requests across selected provider and model targets to optimize reliability and prevent overloading.

You can optimize AI API costs by applying a cost-optimized routing strategy or the 9-factor auto scoring mode. These evaluate target provider and model pairs to route requests to the lowest-cost available option while preserving response continuity.

You need an active OmniRoute endpoint with OMNIROUTE_URL and OMNIROUTE_KEY configured to set up multi-provider routing strategies. The combo definition relies on this endpoint to execute runtime selection, failover, and load balancing.

Routing dry-run simulation validates AI provider selection by testing combo configurations against routing strategies without sending live traffic. This confirms how requests distribute, fail over, and score across targets before production deployment.

Available routing strategies for managing AI provider traffic include priority, weighted, round-robin, least-used, cost-optimized, and context-aware options. A production-friendly auto-combo mode also uses 9-factor scoring to evaluate and select targets.