OllamaR is a high-performance, easy-to-configure open-source load balancing server designed specifically for Ollama workloads. It distributes requests efficiently across multiple backend Ollama servers to improve availability, response speed, and resource utilization. Key features include high-performance request distribution algorithms, straightforward configuration, dynamic scaling through adding or removing backend nodes without restart, and enhanced security by keeping the native Ollama port unexposed and blocking dangerous endpoints related to model and data manipulation. OllamaR acts as a proxy in front of Ollama servers, storing the backend server list in a local database populated via the /api/tags interface, and supports both self-hosted and public Ollama servers. Supported endpoints include root status, /api/tags and /v1/models for listing enabled models, /api/version for version info, /api/chat for LLM inference, and /api/embed and /api/embeddings for embedding generation. Dangerous administrative