The llm CLI is much more unixy, letting you pipe data in and out easily. It can use hosted and local models, including ollama.