Stories with one action


(Sibtain Raza Jamali) #1

Dear Rasa Core Developers, I want to create a chat bot in which there are 120 intents.To perform an action against each intent i have created a story(with only one action) for each intent. So there are 120 unique stories. There is a fall back policy as well. NLU model works perfectly. But when it comes to perform actions against a command, rasa core model some times gets stuck like response=agent.handlemessage(message) gives response with len(response)=0 . There is no problem with domain file. When I tried too debug the stories the same thing happened and bot predicts no action instead of (action_listen). I am using KerasPolicy() and fall back policy and no memoization policy because i don’t want to keep track of previous history. I want bot to perform each action independently(no matter what happened before). Kindly help me.

(Deepak Shetty) #2

Sounds to me that you dont need rasa-core. If you still want to use it - the Keras policy too does use LSTM so youll have to write your own one if you don’t want it to remember anything. Finally you might want to just action_restart after every action so that it doesn’t remember anything.

However I’d suggest you see why you feel you need Rasa-Core because it doesnt really look like you want a dialogue and if you are for e.g. just doing a stateless FAQ answering bot then NLU might be enough for you.