Same input to rasa, detect the same intent, but not predict the same action

I send the same input to rasa twice.

The first time, rasa works fine, and predict action by AugmentedMemoizationPolicy.

But then I send the same input to rasa again, it can not predict by any policies, and result in action_default_fallback.

I debug it, I find this in log: “rasa.core.policies.memoization - There is no memorised next action”

My Rasa version is 2.7.1, I have tried each of the config, but both not work.

  • name: MemoizationPolicy max_history: 5


  • name: AugmentedMemoizationPolicy max_history: 5

Somebody can help me? I am new of rasa, I have no idea about the problem

Is there anybody can give me some clues?

hi @ronan_ly - this should indeed work with the AugmentedMemoizationPolicy. To debug further, you could try setting the max_history to 1.

But more generally, it sounds like you always want this intent to receive the same response, no matter what happened in the earlier parts of the conversation? If so, you can use a rule instead.

Thank you very much, your suggestion helps me I add rules, it works fine!