Don't draw __start__ and __end__ nodes of subgraphs with conditional edges

[amanlai]

Right now (langgraph 0.2.50), if I draw a graph with a subgraph that has conditional edges (that iterates back to the entry node of the subgraph), I get an image where __start__ and __end__ nodes of the subgraph are explicitly drawn. While I understand that there are cases where this is useful for clarity, I think it might be useful to have the option to not draw these nodes, similar to how it's currently drawn if the subgraph does not contain such conditional edges.

I want to be able to draw the graph on the right hand side without the circled nodes similar to the one on the left.

For reproducibility, the following code produces the graph in the right:

from typing import TypedDict, Literal
from langgraph.graph import StateGraph

class SubGraphState(TypedDict):
    value: str

class GraphState(TypedDict):
    val: str

def subfunc(state: SubGraphState):
    return state

def func(state: GraphState):
    return state

def should_continue(state: GraphState) -> Literal["subgraph", "node"]:
    return "subgraph" if state else "node"

subgraph = StateGraph(SubGraphState)
subgraph.add_node("head", subfunc)
subgraph.add_node("hand", subfunc)
subgraph.add_edge("__start__", "head")
subgraph.add_conditional_edges("head", lambda x: "hand" if x else "__end__")
subgraph.add_edge("hand", "head")
subgraph = subgraph.compile()

graph = StateGraph(GraphState)
graph.add_node("router", func)
graph.add_node("node", func)
graph.add_node("subgraph", subgraph)
graph.add_edge("__start__", "router")
graph.add_conditional_edges("router", should_continue)
graph.add_edge("node", "__end__")
graph.add_edge("subgraph", "__end__")
graph = graph.compile()
graph.get_graph(xray=1).draw_mermaid_png(output_file_path="graph.png")

If we replace

subgraph.add_conditional_edges("head", lambda x: "hand" if x else "__end__")
subgraph.add_edge("hand", "head")

with:

subgraph.add_edge("head", "hand")
subgraph.add_edge("hand", "__end__")

in the above code, we get the graph on the left.