Open
Description
I'm using the interrupt() function in agent-chat-ui to enable human interaction during agent workflows. However, I've noticed an unexpected behavior:
When interrupt() is triggered, the previous messages (including tool messages or assistant messages) of the subgraph the interrupt belongs to temporarily disappear from the frontend.
Once the user submits their input, the previously hidden messages reappear.
Is this behavior intentional or a bug?
If intentional, is there a way to configure interrupt() so that it preserves the full visible message history throughout?
Thank you so much!
Metadata
Metadata
Assignees
Labels
No labels