sidecar

An OpenAI-compatible LLM proxy that routes requests to the fastest healthy provider. It runs as a small sidecar, supports latency/error-aware routing, session affinity, and Vertex AI, and can be deployed as a single static binary or Docker container.