langgraph

What problem does it solve?

LangGraph provides a low-level orchestration framework for building stateful, multi-actor LLM applications with cyclic graphs. It enables developers to model state, nodes, and edges with explicit control over execution, persistence, and interaction patterns such as human-in-the-loop.

Core Features & Use Cases

• StateGraph with a defined state schema: Use TypedDict or Pydantic to define the graph state and reducers to manage updates.
• Nodes and Edges: Implement node functions and transitions, enabling conditional routing, fan-out/in, and loops.
• Durable execution & persistence: Leverage checkpointers to persist state across steps and interruptions, supporting long-running agent workflows and interruption handling.
• Subgraphs & composition: Build modular graphs and compose them into larger workflows for scalable architectures.

Quick Start

Create a simple StateGraph with a basic State, add a couple of nodes and a fixed edge from START to END, compile the graph, and invoke it with an initial state to see how updates propagate through the graph.