Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/discoposse/GUIness/llms.txt

Use this file to discover all available pages before exploring further.

Internal graphs let you nest a full pipeline inside a single node. Instead of exposing every step of a complex workflow on the main canvas, you encapsulate the logic inside a custom skill, giving it a clean single-node interface while keeping all the implementation detail accessible when you need it. This is how GUIness supports reusable, composable skills that can themselves be wired into larger pipelines.

Opening an internal graph

Double-click any custom skill node on the canvas to dive into its internal graph. The canvas zooms into the sub-graph view, and a breadcrumb trail appears at the top of the canvas showing your current navigation depth — for example, Main Pipeline › Research Skill. The sub-graph is a full pipeline canvas: you can add, connect, and configure nodes exactly as you would at the top level.

Graph I/O nodes

Sub-graphs communicate with the outside world through two dedicated system nodes:
  • ▶ GRAPH INPUT — Place this node inside the sub-graph to receive data from the parent node’s input ports. It acts as the entry point for anything passed into the custom skill from the outer pipeline.
  • ◀ GRAPH OUTPUT — Place this node inside the sub-graph to send results back out to the parent node’s output ports. Whatever you connect to GRAPH OUTPUT is what the skill exposes to downstream nodes in the outer pipeline.
Click any link in the breadcrumb trail to navigate back up through the graph hierarchy. Clicking the root-level breadcrumb returns you to the main pipeline canvas.
Internal graphs are the primary mechanism for building reusable, composable skills. Define the logic once inside a skill’s internal graph, then drop that skill into any pipeline and wire it up like any other node. Changes to the internal graph propagate everywhere the skill is used.
Primitive nodes — TEXT, INPUTS, COMPUTE, CODE, ROUTER, and CONTEXT — do not have internal graphs. Only custom skills support the double-click-to-enter behavior. Double-clicking a primitive node body on the canvas has no effect.

Build docs developers (and LLMs) love