About this project

LLMRouter is a self-hosted Large Language Model (LLM) API gateway that allows users to unify calls to multiple AI service providers (such as Google, Groq, Mistral, Zhipu AI, Alibaba Cloud Bailian, etc.) locally through a single OpenAI-compatible interface. All keys and configurations are encrypted and stored in a local SQLite database to ensure privacy and security. The system supports automatic failover (switching models upon 429/5xx errors), intelligent routing (automatically selecting the optimal model when model="auto"), streaming output, and tool calling. A Web management panel is provided for key management, routing priority adjustments, and Playground testing. It supports one-click deployment on Windows/macOS/Linux, with idle memory usage of approximately 40MB, making it suitable for individuals or small teams. The technical architecture is based on an Express backend and React frontend, with a modular design for vendor adaptation that is easy to extend.