mirror of
https://github.com/langchain-ai/langgraph.git
synced 2026-09-30 05:25:05 +02:00
Elior Nataf Lackritz
314cb19da8
The six state methods each resolved the saver themselves, and only get_state recast a checkpointer=True subgraph's namespace. So get_state_history and update_state given the subgraph task's config looked under `name:<task_id>`, where the run never stores anything: the history came back empty and the update was silently lost. The state methods and the run loop now share one saver lookup and one namespace rule. The hydration guard also raises when a saver comes without a config, which skipped the replay the same way.