langgraph-master

LangGraph enables designing AI agents through explicit graph-based architectures using nodes and edges to define control flow. The Skill provides architecture patterns, memory management strategies, tool integration approaches, and practical implementation guidance for building type-safe, composable agent workflows.

LangGraph supports six major agent patterns: basic routing, parallel execution, subgraph composition, evaluator–optimizer loops, human-in-the-loop processes, and tool integration. Each pattern addresses different control flow requirements, from simple sequential flows to complex multi-agent hierarchies with checkpointing and streaming.

State management in LangGraph uses explicit state definitions passed between nodes and edges. The Skill covers memory management strategies including checkpointer and store usage, enabling persistent state across agent execution, tool interactions, and multi-turn conversations.

Yes, LangGraph supports parallel node execution, multi-LLM model guidance, and the Command API for advanced routing. The Skill provides concrete examples for orchestrating multiple language models, conditional branching, and concurrent task execution within a single graph.

Tool integration in LangGraph uses dedicated tool nodes connected via edges in the graph. The Skill covers tool node implementation, Command API usage for dynamic tool selection, and patterns for handling tool outputs and error states within agent workflows.

Subgraph composition allows nesting complete graphs as nodes, enabling modular agent design and code reuse. The Skill demonstrates human-in-the-loop patterns where agents pause for user input, approval, or feedback before continuing execution.