Decision Models Need a Router, Too
One API for a growing decision-model ecosystem. Provider logos illustrate the ecosystem; the experiment below measures Kai → Vega.
Save your largest decision model for the requests that need it.
vLLM Semantic Router already routes across open and closed LLMs. Decision models are becoming a model pool of their own. Open families such as Decision 2.0, Clef, Decider and Laya sit alongside hosted APIs such as Jev and GLiDE. They classify intent, score candidates and check conditions—but which model should handle each request?
Meet System One Auto. Connect self-hosted decision models or compatible hosted APIs behind vllm-sr/auto. Configure how to select a model, when to escalate through a cascade, and whether an LLM judge should review the final candidates. Your application keeps one native API; you choose the model pool and routing policy.