LangGraph 0.1.0 Migration Guide: Breaking Changes and Fixes
LangGraph 0.1.0 introduces critical breaking changes for developers, including deprecated functions, Node.js version updates, and build process...
Direct answer
LangGraph 0.1.0 introduces breaking changes such as the deprecation of `createReactAgent` (replaced by LangChain's `createAgent`), removal of `toLangGraphEventStream`, and a requirement for Node.js 20+. Developers must update their code to use these new functions, ensure their environment meets the Node.js version, and adjust imports if using the `dist/` directory.
LangGraph, the LLM workflow orchestration framework from the LangChain team, has released version 0.1.0 with several breaking changes that developers must address to ensure their applications continue to function correctly. This guide outlines the key changes and fixes to help teams migrate smoothly.
One of the most significant updates is the deprecation of the `createReactAgent` function. Developers should now switch to LangChain's `createAgent` function, which runs on LangGraph and includes a flexible middleware system to enhance agent functionality.
LangGraph 0.1.0 also enforces a stricter Node.js requirement: all packages now need Node.js 20 or higher, dropping support for Node.js 18. This decision aligns with Node.js 18 reaching its end-of-life in March 2025.
Other key changes include the removal of the low-level `toLangGraphEventStream` helper function—stream responses are now handled by the SDK, and low-level clients can select wire formats via encoding options passed to `graph.stream`. Additionally, the `langgraph` package’s build process has shifted to a bundler-based approach instead of raw TypeScript output; those importing from the `dist/` directory (not recommended) will need to update their imports to use the new module system. Finally, developers can now define interrupt types during graph construction to strictly type values passed to and from interrupts.
For context, LangGraph is designed to handle agent loops, state management, and tool calls efficiently. Its core concepts include nodes (minimal execution units like LLM calls or tool use), edges (flow rules between nodes), and state (a global shared storage for intermediate results).
Sources
- LangGraph v1 Migration Guide: https://python.langchain.com/docs/guides/langgraph/migration
- LangChain Official Documentation Index: https://docs.langchain.com/llms.txt
- LangGraph v1 Release Notes: https://python.langchain.com/docs/guides/langgraph/whats_new
FAQ
- What replaces the deprecated `createReactAgent` function in LangGraph 0.1.0?
- Developers should use LangChain's `createAgent` function, which runs on LangGraph and includes a flexible middleware system for enhanced agent capabilities.
- Why does LangGraph 0.1.0 no longer support Node.js 18?
- Node.js 18 reached its end-of-life in March 2025, so LangGraph 0.1.0 has dropped support to align with security and maintenance best practices.
- What happened to the `toLangGraphEventStream` helper function?
- The low-level `toLangGraphEventStream` function has been removed; stream responses are now handled by the SDK, and low-level clients can select wire formats via encoding options in `graph.stream`.
- How has the `langgraph` package's build process changed in 0.1.0?
- The package now uses a bundler-based build approach instead of raw TypeScript output. If you import from the `dist/` directory (not recommended), you'll need to update imports to the new module system.