langgraphlisted

# LangGraph Patterns LangGraph is a framework for building stateful, multi-step workflows with LLMs. It models workflows as directed graphs where nodes are functions and edges define control flow. Use it when you need conditional routing, parallel execution, human approval gates, or persistent state across workflow steps. ## Quick Reference | Pattern | Description | Reference | |---------|-------------|-----------| | State Management | TypedDict/Pydantic schemas, reducers, MessagesState | [references/state-management.md](${CLAUDE_SKILL_DIR}/references/state-management.md) | | Conditional Routing | Branching, retry loops, quality gates, END fallbacks | [references/routing.md](${CLAUDE_SKILL_DIR}/references/routing.md) | | Parallel Execution | Fan-out/fan-in, Send API, map-reduce, error isolation | [references/parallel-execution.md](${CLAUDE_SKILL_DIR}/references/parallel-execution.md) | | Human-in-the-Loop | Interrupt points, approval gates, feedback loops, API integration | [references/human-in-loop.md](${CLAUDE_SKILL_DIR}/references/human-in-loop.md) | | Checkpoints | MemorySaver, PostgresSaver, Store, crash recovery, debugging | [references/checkpoints.md](${CLAUDE_SKILL_DIR}/references/checkpoints.md) | | Supervisor | Coordinator agents, round-robin/priority dispatch, LLM-based routing | [references/supervisor.md](${CLAUDE_SKILL_DIR}/references/supervisor.md) | ## Pattern Summaries ### State Management Design workflow state with TypedDict (lightweight, internal) or P