nodes naming convention

[jonathanbouchet]

Hello,
I just started with LangGraph and I may have a naive question regarding naming nodes.
So far I really like LangGraph.

I'm using

• langGraph 0.0.69
• langchain-core 0.2.9
• Python 3.11.8

I have this simple graph with 2 nodes that calls a different function:

def greeting(input: str) -> str:
    return f"Hello {input}"

def checking(input: str) -> str:
    return f"{input}, how are you ?"

workflow = Graph()

workflow.add_node("greeting_node", greeting)
workflow.add_node("checking_node", checking)

workflow.add_edge("greeting_node", "checking_node")
workflow.add_edge("checking_node", END)

workflow.set_entry_point("greeting_node")
# workflow.set_finish_point("checking_node")

app = workflow.compile()

This gives the following graph as expected

However when I replace the names of the nodes with :

workflow2 = Graph()

workflow2.add_node("node_1", greeting)
workflow2.add_node("node_2", checking)

workflow2.add_edge("node_1", "node_2")
workflow2.add_edge("node_2", END)

workflow2.set_entry_point("node_1")

app2 = workflow2.compile()

it looks confused by naming nodes starting with the same word:

Is it advised to have a completely different names to avoid this regex issues ?

thanks in advance.

[hwchase17]

huh this seems like a bug, we'll look into it

[hinthornw]

Just to check, are you using Graph or StateGraph for this?

[jonathanbouchet]

I'm using Graph

[jonathanbouchet]

I did a quick test using StateGraph

from typing import TypedDict, List

class AgentState(TypedDict):
    pass

workflow2 = StateGraph(AgentState)

workflow2.add_node("node_1", greeting)
workflow2.add_node("node_2", checking)

workflow2.add_edge("node_1", "node_2")
workflow2.add_edge("node_2", END)

workflow2.set_entry_point("node_1")

app2 = workflow2.compile()

I get the same result as using Graph

[jonathanbouchet]

a quick update.

• I have retried with the latest version of LangGraph : '0.1.1': it produces the same result as initially.
• I also tried with names having different number of characters but the same start: node_1 (still using '0.1.1')

def greeting(input: str) -> str:
    return f"Hello {input}"

def checking(input: str) -> str:
    return f"{input}, how are you ?"

workflow = Graph()

workflow.add_node("node_1", greeting)
workflow.add_node("node_12", checking)

workflow.add_edge("node_1", "node_12")
workflow.add_edge("node_12", END)

workflow.set_entry_point("node_1")

app = workflow.compile()

This time there's no issues.

[jonathanbouchet]

sorry for spamming: one last test: added comments in the functions:

def greeting(input: str) -> str:
    print(f"in greeting: {input}")
    return f"Hello {input}"

def checking(input: str) -> str:
    print(f"in checking: {input}")
    return f"{input}, how are you ?"

gives the expected result:

app.invoke("john doe")

# in greeting: john doe
# in checking: Hello john doe
# 'Hello john doe, how are you ?'

So the issue is only happening when displaying the graph, and not with langGraph itself.