​
Expanding and collapsing runchat nodes

You can publish any runchat workflow to a library and then use it as a node within some other workflow. We call these nodes Tools. After you run a Tool in a workflow, the internal state of the tool (all input and output parameters) will be saved on that node. You can then view this internal state (called a subflow) to see exactly how data flowed through the logic of the tool by clicking the ArrowsOut icon in the node settings bar. You can collapse the workflow back into a single node by pressing the button again.
The expand and collapse toggle will be disabled until you run the runchat and create some internal state to view.

​
Expanding and collapsing Agent nodes

You can also expand Agent node to view the internal state of any Tools used by that node to return a response to your prompt. First, add tools to the Agent node in the node settings bar, then run the node. If a tool was used you will then be able to expand the Agent node using the ArrowsOut button in the settings bar.

​
Internal state

The expanded view shows the internal state of the tool when it was last run. If you run the parent node (Agent or Runchat) again, the inputs and outputs of the child nodes will update to reflect this new state.This is a useful diagnostic tool for detecting bugs in workflow logic or flaws in prompts.

​
Editing subflows

You can make changes to nodes in a subflow (e.g. to Code or Agent nodes) and these will be saved as a custom override to the Tool. This is useful for quickly exploring variations on prompts or fixing bugs.