Ollama runs locally. What's the best option for calling the new Mixtral model on someone else's server programmatically?