What causes the D⇢G and G→F edges in this graph?

[kinghuang]

I created a graph with two nested graphs. When I generate a Mermaid diagram, I'm seeing extra edges D⇢G and G→F that I'm not expecting. Am I doing something wrong with my graph construction? I haven't added any edges or conditional edges for those paths, as far as I know.

Here's the diagram and code (names have been obfuscated).

def create_graph():
    graph = StateGraph(
        state_schema=GraphState,
        config_schema=GraphConfigSchema,
        input=GraphInputState,
        output=GraphOutputState,
    )

    graph.add_node("A", things.A)
    graph.add_node("B", _create_first_subgraph().compile())

    graph.add_edge(START, "A")
    graph.add_conditional_edges(
        "A",
        partial(_send_over_state_list_values, state_key="things", target_node_name="B"),
        then=END,
    )

    return graph


def _create_first_subgraph():
    graph = StateGraph(
        state_schema=GraphState,
        config_schema=GraphConfigSchema,
        output=SubGraphOutputState,
    )

    graph.add_node("C", things.C)
    graph.add_node("D", things.D)
    graph.add_node("E", _create_second_subgraph().compile())
    graph.add_node("F", things.F)
    graph.add_node("G", things.G)

    graph.add_edge(START, "C")
    graph.add_edge(START, "D")
    graph.add_conditional_edges(
        "D",
        partial(_send_over_state_list_values, state_key="things", target_node_name="E"),
        then="F",
    )
    graph.add_edge("F", "G")
    graph.add_edge("G", END)

    return graph


def _create_second_subgraph():
    graph = StateGraph(
        state_schema=GraphState,
        config_schema=GraphConfigSchema,
        output=SubGraphOutputState,
    )

    graph.add_node("H", things.H)
    graph.add_node("I", things.I)

    graph.add_edge(START, "H")
    graph.add_conditional_edges(
        "H",
        lambda state: "continue" if state["things"][0].score != 0.0 else "end",
        {"continue": "I", "end": END},
    )
    graph.add_edge("I", END)

    return graph


def _send_over_state_list_values(
    state: GraphState, config: RunnableConfig, state_key: str, target_node_name: str
):
    values = state[state_key]  # type: ignore [literal-required]

    return [Send(target_node_name, {**state, state_key: [value]}) for value in values]

[NickSherrow]

Hey @kinghuang, I have the same issue (connections between nodes that are not stated anywhere in the graph). Have you figured it out yet? And have you had a chance to check if it affects the outputs, or just a bug in Mermaid?

[kinghuang]

It hasn't affected the output in my graphs. Just a weird artifact in the generated Mermaid diagrams.