[Python] LangGraph: How to determine which node triggered the current node during execution?

Hello,


I want to know which previous node activated the current node when it runs — in order to visualize which edge is active, not just at any given moment, but in all types of graph structures, including conditional branches, parallel flows, and complex multi-path executions.

Here are my trials:


get metadata from

graph.astream_event()


I printed metadata but only got information about current node like langgraph step, node name, checkpointns etc. as I can see from langgraph/libs/langgraph/langgraph/pregel/algo.py

metadata = {
"langgraph_step": step,
"langgraph_node": name,
"langgraph_triggers": triggers,
"langgraph_path": task_path,
"langgraph_checkpoint_ns": task_checkpoint_ns,
}


Also I tried

I also tried manually passing a custom node_history (a list like ['node1', 'node2', ...]) via dispatch_custom_event. This worked for simple graphs, where the execution path is linear or unambiguous.

However, in more complex graphs, like this:

example of complicated graph

...this approach becomes unreliable. For example:

If both B1 and B2 execute and lead to C2, I can’t tell which one triggered it.

If multiple nodes (C1, C2, D) can lead to E, it's unclear which one actually caused E to run.

Even when I tried combining the metadata (e.g., langgraph_step) with my custom node_history, there were always edge cases that made the results ambiguous.

To conclude, what I want is to know which edge is going on at the time. For that, I need which node activated current node. but it seems there is no easy way for this.

My Goal


I want to know which edge is currently active — in other words, which node triggered the current one. But so far, it seems there’s no straightforward way to get this information.

Question


Has anyone found a clean solution to this? Is there a built-in way to get the "previous node" or the incoming edge that activated the current node?

Thank you, developers all around the world — you're the best!

Continue reading...