Skip to content
Adagio

Working with Nodes

Nodes are the working units of a pipeline. Most nodes represent one plugin action.

Adding a node

  1. In the action tray, browse by plugin or search by name.
  2. Click the action to place it on the canvas, or drag it to a specific position.

Node anatomy

Each node has:

  • Name: the action label shown on the canvas
  • Input ports: typed values the action consumes
  • Output ports: typed values the action produces
  • Parameters: configuration values that are fixed on the node or read from a variable

Connected local-plugin nodes also show Refresh <plugin> interface. Refresh is available only for a private local plugin backed by a usable Conda environment. Starting it from one node refreshes every action and node from that plugin.

Connecting nodes

Drag from an output port to an input port to create a connection. Adagio enforces type compatibility, so a port only accepts compatible outputs.

If you want help discovering the next valid step, select a port first. The action tray will filter to compatible actions.

Configuring parameters

Select a node to open it in the inspector.

You will usually see three kinds of parameter behavior:

  • Required with no default: you must set a value on the node or use a variable
  • Optional with an action default: you can keep the plugin’s default, override it with a value on the node, or use a variable
  • Metadata column parameters: these work with a metadata input plus a selected column name

Use these rules:

  • Set a value on the node when it should always be fixed in the pipeline. This is also the quickest way to iterate in the editor.
  • Leave Use default enabled when the plugin default is the right fixed behavior.
  • Use a variable, with the variable button at the right of the parameter field, when several parameters should share one value or the runner should be able to change it.

See Inputs, Parameters, and Defaults for the full model.

Metadata-aware actions

Some actions take metadata directly. Others take a metadata file plus a separate metadata-column parameter such as MetadataColumn[Categorical].

In those cases, the connection carries the metadata object and the parameter identifies which column to use.

Removing a node

Select the node and remove it. Deleting a node also removes:

  • its edges
  • any pipeline inputs that existed only to feed that node

Variables the node used stay in the Variables tab. Delete them there if nothing else needs them.