Used to allow small LLMs to to make intention decisions and then dynamically switch between larger LLMs for different tasks. Also allows override based on document types. For example in the even of an image, it will use llava automatically. I will be adding user valves in a future revision to allow easier swapping of models via the UI.