vllm-project / semantic-router

Why you might care

Useful if you run agents or apps across several models or backends (local, private, cloud) and want a policy to pick the model per call for quality, latency, cost, or data-location constraints. It sits in front of existing inference endpoints, so the agent harness keeps its own loop and tools.