You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I have a two tier hierarchy of graphs with some shared state between the "master" and the "subgraphs".
Each subgraph shares a partial set of state from the master in order to access the relevant current state when initiated, mutate it and return it back to the master.
The issue I'm seeing is that streaming updates seem to include state that was unchanged neither by the master or the subgraphs:
for namespace, type, chunk in supervisor_agent_graph.stream({ "messages": HumanMessage(request.message)}, config, stream_mode=["custom", "updates"], subgraphs=True):
if type == "updates":
[(node_name, value)] = chunk.items()
#value includes state that was never changed
last_update_value = value
Could it be that the "copying" of state forwards to the subgraph and back to the master is considered a state "update" even when the actual value of the state field does not change and is not updated explicitly by a node in the subgraph?
The text was updated successfully, but these errors were encountered:
I have a two tier hierarchy of graphs with some shared state between the "master" and the "subgraphs".
Each subgraph shares a partial set of state from the master in order to access the relevant current state when initiated, mutate it and return it back to the master.
The issue I'm seeing is that streaming updates seem to include state that was unchanged neither by the master or the subgraphs:
Could it be that the "copying" of state forwards to the subgraph and back to the master is considered a state "update" even when the actual value of the state field does not change and is not updated explicitly by a node in the subgraph?
The text was updated successfully, but these errors were encountered: