langgraph with stream

[goodchengwang]

this is my code

  const InputStateAnnotation = Annotation.Root({
    question: Annotation<string>,
  })

  const StateAnnotation = Annotation.Root({
    question: Annotation<string>,
    context: Annotation<Document[]>,
    answer: Annotation<string>,
  })

  // Define application steps
  const retrieve = async (state: typeof InputStateAnnotation.State) => {
    const retrievedDocs = await vectorStore.similaritySearch(state.question)
    return { context: retrievedDocs }
  }

  const generate = async (state: typeof StateAnnotation.State) => {
    const docsContent = state.context.map((doc) => doc.pageContent).join('\n')
    const messages = await ragPrompt.invoke({
      question: state.question,
      context: docsContent,
    })
    const response = await model.invoke(messages)
    return {
      answer: response.content,
    }
  }

  // Compile application and test
  const graph = new StateGraph(StateAnnotation)
    .addNode('retrieve', retrieve)
    .addNode('generate', generate)
    .addEdge('__start__', 'retrieve')
    .addEdge('retrieve', 'generate')
    .addEdge('generate', '__end__')
    .compile()
  const stream = await graph.stream(
    {
      question: query,
    },
    {
      streamMode: 'messages',
    },
  )

  return LangChainAdapter.toDataStreamResponse(stream)

but when i run this code, it will throw error with Error in handler StreamMessagesHandler, handleLLMNewToken: TypeError [ERR_INVALID_STATE]: Invalid state: Controller is already closed Can anyone help me?

[dithmal]

hey, were you able to solve this? I'm getting the same error.

[github-actions[bot]]

This discussion was automatically locked because it has not been updated in over 30 days. If you still have questions about this topic, please ask us at community.vercel.com/ai-sdk