Testing LangGraph with static interrupts – best practices?

OSS Product Help LangGraph
1

Hi all,

I’m testing a LangGraph workflow with several conditional branches and LLM nodes.

To do this, I’m using static interrupts as described in the docs:

My current approach:

  • Add static breakpoints at key nodes
  • Use graph.get_state(thread) to assert state before/after node execution
  • Resume with graph.ainvoke(None, thread)
  • Write one test per conditional path

Questions:

  • Has anyone used this approach for testing?
  • Any tips or common pitfalls?
  • Is there a better way to test graph logic at this level of detail?

Appreciate any advice!