What breaks in document AI

Runs lost on crash, resume fails

18 of 18 quotes

• “crons fail after a restart”

• “The case we hit is a human-in-the-loop pause.”

• “Workaround we run today”

• “2 runs (which take longer than 20-30 min) were cancelled and resumed multiple times and resulted in failed state due to exceeded max attempts (3) .”

• “recently we are experiencing some issues related to some of our graph runs getting cancelled (asyncio. CancelledError ) in the LangSmith tracing in our Production deployment.”

• Hit this one too? We are building for it. Join the waitlist.

• “Determinism sucks, you do have to work hard and make everything idempotent in activities like we would for durable software anyway.”

• “now see duplicate log lines on every resume.”

• “Interrupt() in a loop will cause extra resumes”

• “In the end, we completely abandoned this solution and are now saving the event log ourselves. This is something I expected LangGraph to do, but had to re-implement.”

• “resuming from an interrupt inside the billing subgraph (ie. asking for the order ID) is not working.”

• “Hi @Pradeep ran into the same problem.”

• “The error occurs sporadically in production so I'm unable to create a reliable reproduction.”

• “But in my case, it produces 2 AI messages.”

• “Resuming after interrupt doesn't reuse prior task outputs when interrupt is in subgraph”

• “The custom state updates made via ToolRuntime and Command.update within tools are lost when control returns from the agent to the parent deep agent.”

• “when I restart the conversation, the checkpointed messages are always missing some content”

• “in langgraph cloud its not working when resuming.”

• “the new checkpoint it generates doesn't inherit the next field from the checkpoint_id I passed in and instead gets it from the latest checkpoint of the thread.”

• Hit this one too? We are building for it. Join the waitlist.