Key Concepts

NodeTool workflows are represented as Directed Acyclic Graphs (DAGs). Each node performs a unit of work and passes data to the next node.

• Nodes – Encapsulate a specific operation. Nodes have inputs, outputs and configurable properties.
• Graph – A collection of nodes and their connections. Use graph() to build graphs and run_graph() to execute them.
• DSL – NodeTool provides a Python domain specific language with modules for different domains (nodetool.dsl.chroma, nodetool.dsl.google, …).
• WorkflowRunner – The engine that executes graphs. It handles parallel execution, GPU management and progress updates.
• ProcessingContext – Holds runtime information like user data and authentication tokens.

Understanding these concepts will help you design efficient workflows and build your own nodes and agents.

Node Type Resolution

When a workflow references a node by its type string (e.g., package.Namespace.Class), NodeTool resolves the class using a robust strategy:

• In-memory registry lookup (with and without a trailing Node suffix)
• Dynamic import of modules based on the type path, then re-check the registry
• Lookup in the installed packages registry for external nodes
• Fallback match by class name only, ignoring an optional Node suffix

This enables loading graphs without pre-importing all node modules and supports short class-name references in some cases.