Issue implementing langgraph use case with checkpointer and langserve

[israelberko]

I have the following graph implementation:
I random n number of question and then want to ask them one by one to the user and get a response on each on of it.
the graph stops before the first collect_answer node but I don't manage to resume it once it stops
any idea what I'm doing missing?

def create_graph() -> CompiledGraph:
    workflow = StateGraph(GenerativeUIState)

    workflow.add_node("prepare_questions", prepare_questions)
    workflow.add_node("ask_question", ask_question)
    workflow.add_node("collect_answer", collect_answer)
    workflow.add_edge("prepare_questions", "ask_question")
    workflow.add_edge("ask_question", "collect_answer")
    workflow.add_conditional_edges("collect_answer", has_open_questions)
    workflow.set_entry_point("prepare_questions")  # Set the new entry point
    workflow.set_finish_point("collect_answer")

    memory = MemorySaver()

    graph = workflow.compile(checkpointer=memory, interrupt_before=["collect_answer"])
    # graph = workflow.compile( interrupt_before=["collect_answer"])
    # graph = workflow.compile()
    return graph

and the langserve part:

def start() -> None:
    app = FastAPI(
        title="Edgewise Backend",
        version="1.0",
        description="a very good description is written here",
    )

    # Configure CORS
    origins = [
        "http://localhost",
        "http://localhost:3000",
    ]

    app.add_middleware(
        CORSMiddleware,
        allow_origins=origins,
        allow_credentials=True,
        allow_methods=["*"],
        allow_headers=["*"],
    )

    graph = create_graph()

    runnable = graph.with_types(input_type=ChatInputType, output_type=dict)

    # add_routes(app, runnable, path="/chat", playground_type="chat")
    add_routes(app, runnable, path="/chat", playground_type="default")
    print("Starting server...")
    uvicorn.run(app, host="0.0.0.0", port=8000)

this an image of the playground:

[eyurtsev]

langgraph is not officially compatible with langserve (https://github.com/langchain-ai/langserve?tab=readme-ov-file#%EF%B8%8F-langgraph-compatibility)