Ask HN: For a team experienced with LLMs – Any concrete reason to use LangGraph?
Never used LangChain\LangGraph , saw the bad reviews about LangChain (albeit they are 1+ year old) - has anything changed? What can we do easier\faster with the framework rather than building our own pipeline? What unexpected things pop up, especially during maintenance, debugging and scale? Are there other frameworks you would recommend?
langgraph’s killer for llm-savvy teams needing complex, stateful workflows, think multi-agent systems or cyclic graphs. it’s faster than custom pipelines, with langsmith debugging and one-click scaling. but it’s steep to learn, and debugging multi-agent setups can suck if state’s messy. langchain’s better for simple, linear tasks; langgraph’s for intricate control. autogen’s easier but less precise; crewai’s rigid. langgraph’s your pick for dynamic, production-ready projects.
Not really.
Langchain was designed for chaining requests together, back when you had to maintain context with AI yourself. OpenAI figured a fix that works for both humans and AI - putting comms in a chat format, and so we didn't need to chain input.
Langchain was one of the earliest to RAG, but it was more of a hack and didn't do as well as many alternatives.
It seems like they're going into agents now. No comment but it feels like they're a step behind.
Thanks for the input