How to influence the dialog model prediction based on already identified slots

Rasa Open Source
1 
## Ask location and Show weather
* get_weather
  - utter_ask_location
* get_weather{"location": "New York"}
  - slot{"location": "New York"}
  - utter_city_weather

## Show weather if user provided a location
* get_weather{"location": "New York"}
  - slot{"location": "New York"}
  - utter_city_weather

## Show weather even if the user didn't provide weather in the current utterance, but instead a few messages back in earlier conversation  (Thus assuming the Slot of location is saved)
* get_weather
  - slot{"location": "New York"}
  - utter_city_weather

I have the above outlined Scenario where i have a Slot for location , and first two stories cover the ideal cases where user is asked for a locations before providing the weather. the first two stories are working fine.

But the last story, where if the user already provided location at some point earlier in conversation, and the Slot for location is saved and exists in the Tracker, I want the model to directly predict utter_city_weather without asking for a location because location slot already exists in the Tracker.

I am expecting the last story to capture this scenario, but it’s not working. How do i make it possible?

(Note: I have double checked and location Slot is indeed present in the Tracker, still the third story doesn’t work)

2

@Juste I referred to Designing Stories with Slots section in your blog post of Designing Rasa training stories to model this behaviour. Any idea what am I missing here?

3

Mhm, so this slot event is not being used correctly, a slot event should appear in the story where the slot is set, not where it is checked. Can you post the part of the story where the user provided their location already? ideally you would write a story that encompasses this whole flow, i.e.

* some_other_location_input{"location": "New York"}
  - slot{"location": "New York"}
  - some_other_utterance
* get_weather
  - utter_city_weather

or something of the sort.