A graph is a network of connected nodes and edges. In LangGraph, it becomes a practical map of an AI application: nodes perform actions, edges direct execution, and shared state carries information between steps. section 10.2.1 section 10.2.2
Use these parts to read a LangGraph workflow:
- Nodes represent individual actions that read the current state, modify it, and return updated state. section 10.2.2
- Directed edges connect actions and determine which node executes next. section 10.2.2
- The initial state enters the first node, while the final state leaves the last executed node. section 10.2.2
From relationships to workflow
A social network can model people as nodes and friendships as edges; similarly, a LangGraph application can model actions as nodes and execution relationships as directed edges. One example graph computes the sum of squares of numbers. section 10.2.2
Figure 10.4 In LangGraph, nodes take in the graph state and modify it.
Trace how nodes receive graph state and produce modified state as execution moves through the workflow. section 10.2.2
