Jev Router

Jev Router is an open-source, self-hosted LLM routing layer that picks the right model for every request. Core capabilities: • One OpenAI-compatible endpoint for all your models • Automatic model selection with send model set to "auto" • Strategies for balanced, quality, cheapest, fastest, or privacy routing • Automatic fallbacks when a provider fails • Built-in guardrails and full routing observability • Streaming, tool calling, and JSON mode support Jev Router works as a drop-in replacement for the OpenAI SDK. Point your existing code at the router, change the model name to auto, and keep everything else. Each response reports which model answered and why it was chosen, and the same decision is returned in X-Jev-Router headers so streaming responses are covered too. The router connects to Liquid AI, OpenAI, Anthropic, Google Gemini, DeepSeek, Groq, OpenRouter, Ollama, LM Studio, and vLLM, so you can keep the providers you already use. Because routing happens on your own server, your API keys and prompts stay under your control. Teams can cut costs by sending simple requests to smaller models while reserving frontier models for hard tasks, and the project reports median routing decisions in about 295 ms. Built for developers, platform engineers, and startups who run several LLM providers and want lower spend, fewer failures, and clear visibility into model choices.