Nodes as Actions

Nodes as Actions

Understanding the fundamental unit of work in a graph.

Nodes as Actions

In LangGraph, a Node is where "work" happens.

What is a Node?

A node is simply a Python function.

  • Input: The current State.
  • Process: Do something (call API, run computation, think).
  • Output: An update to the State.
graph LR
    StateInputs[Current State] --> Node((Node Logic))
    Node --> API[Call LLM/Tool]
    API --> Result
    Result --> StateUpdate[State Update]

Code Example

def my_node(state):
    # 1. Read from State
    question = state["messages"][-1]
    
    # 2. Do Work
    answer = call_llm(question)
    
    # 3. Update State
    return {"messages": [answer]}

Think of nodes as meaningful steps in a flowchart. If you would draw a box for it on a whiteboard, it should be a node.

Subscribe to our newsletter

Get the latest posts delivered right to your inbox.

Subscribe on LinkedIn