langgraph/tutorials/usaco/usaco/

[giscus[bot]]

langgraph/tutorials/usaco/usaco/

Build language agents as graphs

https://langchain-ai.github.io/langgraph/tutorials/usaco/usaco/

[jojohannsen]

This would be simpler if human were a node vs. interrupting the graph, then re-entering. What's the reasoning behind this choice?

[hinthornw]

Great question! Interrupting the graph ensures the full state is checkpointed, so the human could respond immediately or in 3 years without impacting the outcome.

You could do that with an explicit human node that interrupts or by routing to the END as well, though I'm not sure if those to options are any simpler.

There's definitely ways we could simplify this notebook though so if you have a more elegant solution I'm always looking to learn!

[jojohannsen]

thank you. The state checkpoint is great, there's a lot of things that makes possible.