It should be if true then require HITL. if false, output directly. The response from AI Agent only showed after i stop the execution. When the execution is running, no response is showed in the logs.
Another thing is that i cant seem to have the settings to set the approval in the “Respond to chat” node.
”when chat message received node” - respond mode: using respond nodes.
Using Response Node is the correct option, if you click on the dropdown menu you will see it’s description before selecting it which details that.
Seems like your prompt is using a lot of tokens or something is hanging somewhere. Could be the model, could be the MCP etc. You can test be switching things off one by one.
Alright, so if I want to use the ‘Respond to Chat’ node as a Human-In-The-Loop (HITL) step in the chat interface, it’s not possible, right? I would need to implement the logic myself using different nodes.
My use case would be: chat input → AI → HITL → user interface prompt, where the user can click ‘Yes’ or ‘No’.
Yes, this is achievable, but still you won’t need a respond to chat node. What you would do is something like this:
Chat (respond on last node) -> AI -> Notify human & capture response -> process if needed in the last node
Also, one thing we forgot to test while you were still using the Respond chat node is disabling Wait for user Input. You can try again with this disabled, as it waits for another message from the user, which is not always needed.
