It just seems to me that Node-Red is good for describing flows and is easy to use graphically. GStreamer is also based on flows, though it doesn’t have a great graphical description environment.
Now, Node-Red is written in JavaScript and is all about sending messages. The nodes in Node-Red are static in general, the same as GStreamer, so it might be possible to analyze the layout of the Node-Red flowgraph at launch time in order to call gst-launch. GStreamer plugins also support messages between each other, often via GstBus.
Where I think this is important for Beagle is around users new to AI. I find the Python-based tools in the Edge AI SDK can be a bit hard for new users to understand and visualize in ways that let them customize easily.
Thoughts?