node-graph-guidelisted

# Node graph Guidelines (Visual Data-Flow Authoring) Engine-agnostic architecture for a visual node-based graph that authors and non-programmers use to assemble behavior by wiring typed nodes together, then compile or evaluate into executable work. The graph itself is plain typed data (see data-model-guide); its node arrays are laid out for traversal (see data-oriented-design-guide); when it emits render/compute work, the scheduling of that work belongs to gpu-rendering-guide. ## Requirements - A reflectable typed object/data store the graph serializes into (nodes, connections, per-node settings, interface), so the editor, evaluator, and tools all read the same definition, see data-model-guide. - A node-type registry that plugins can extend with new node kinds and new wire data types without modifying the core. ## Essentials - **Graph is data, not code** - Nodes, typed pins, wires, and constant inputs are serialized objects, not topology hidden in imperative code, see [references/node-model.md](references/node-model.md) - **Typed pins gate every wire** - A wire is legal only when the producer's output type satisfies the consumer's input type, see [references/typing-and-connections.md](references/typing-and-connections.md) - **Compile, then run** - Lower the authored graph into a flattened, ordered executable form and evaluate that, instead of walking the editor graph every frame, see [references/evaluation-and-compilation.md](references/evaluation-and-compilation.md) - *