Browser issues using interactive train in Rasa X

[amsolu]

Which is the best browser to work with Rasa-X?

I was using chrome, and sometimes depending on the conversation and the use of buttons when I am reaching the end and the conversation screen have to scroll up, it is overlapping the template messages with the next actions and I can no push the buttons. I am attaching the pic for better understanding

When I select the action, the text template only is displayed (no buttons) Then when I select the “action_listen” as next action, then the buttons are displayed but I can not select any because they are overlapped with the action_listen.

Any suggestion, recommendation, workaround?

Thanks

[amsolu]

Hi @erohmensing or @Juste, Any idea if this is a browser issue? How can fix this behavior?

[erohmensing (Ella Rohm-Ensing)]

Hi @amsolu, this is probably not a browser issue, but rather probably a front-end bug. I think our front-end engineers are already aware of this, but you can make sure they are and track the problem by creating a github issue for the bug.

[ravisurapati]

@amsolu i am having one doubt, please help me, its urgent

how r u i am not able to get something working with rasa if i have to ask bot with some question, it reply me with " what about coffee…? "

1. yes 2. no 3 . notyet

1.and if i type yes i will get answer for yes 2.if i type no, i will get answer for no 3.if i type notyet, will get answer for that

answer me please, i am using rasa x latest version please answer me with

domain.yml and stories.md file also, please… need urgent

[ravisurapati]

i am having one doubt, please help me, its urgent

how r u i am not able to get something working with rasa if i have to ask bot with some question, it reply me with " what about coffee…? "

1. yes 2. no 3 . notyet

1.and if i type yes i will get answer for yes 2.if i type no, i will get answer for no 3.if i type notyet, will get answer for that

answer me please, i am using rasa x latest version please answer me with

domain.yml and stories.md file also, please… need urgent

[erohmensing (Ella Rohm-Ensing)]

Hi @ravisurapati. Unfortunately I can’t provide you with stories and domain files at the moment, but can explain them to you. You can implement this by having 3 intents (affirm, deny, not_yet). Then in your domain file create an utterance like utter_ask_coffee with the text as “what about coffee?” And then buttons with the title/payload combos (“Yes”, "/affirm“), (“No”, “/deny”), (“Not Yet”, “/not_yet”). The slash means that when they clock the button it will be immediately classified as that intent. Your stories then would have different paths for the 3 options – the moodbot example that’s in our tutorial in the docs has a good example of button yes/no answers influencing the conversation path.

[ravisurapati]

thanks for the answer, but i want one thing to know that, i have to create three stories for that to get…? and how to train the bot in interactive learning mode, even i am not getting slots in interactive mode

[ravisurapati]

@erohmensing , when i am training the bot in interactive training mode, i am not able to get the data exactly, actions will not show me the proper action to get

[erohmensing (Ella Rohm-Ensing)]

This story doesn’t seem like it should have any slots. If you want to store the yes, no, not yet in a slot, (though that seems unneeessary) this should be handled by your stories. If you don’t have stories for them yet, i’d recommend doing this in interactive learning.

actions will not show me the proper action to get

If it is not predicting the correct actions, then you need to edit the actions to pick the correct one, that is what interactive mode is for. I.e. where the bot predicted action_listen after /not_yet, that should be corrected to say the bot should have performed some other action.

[amsolu]

Hi, Another issue when I’m using browser is the timeout ERROR @erohmensing did you get this or could you point out to what is generating this message?

Training Core model...
Core model training completed.
Training NLU model...
[2019-06-12 21:33:43 -0400] [19713] [ERROR] Exception occurred while handling uri: 'http://localhost:5002/api/projects/default/models/jobs'
Traceback (most recent call last):
  File "/home/amvim/anaconda3/envs/rx1/lib/python3.7/site-packages/sanic/app.py", line 917, in handle_request
    response = await response
  File "/home/amvim/anaconda3/envs/rx1/lib/python3.7/site-packages/rasax/community/api/decorators.py", line 184, in decorated_function
    return await await_and_return_response(args, kwargs, request)
  File "/home/amvim/anaconda3/envs/rx1/lib/python3.7/site-packages/rasax/community/api/decorators.py", line 114, in await_and_return_response
    response = await response

Thank you

[erohmensing (Ella Rohm-Ensing)]

@amsolu this is strange. This is happening when you use the train button on rasa x right? does the same issue happen if you use the CLI command rasa train? It also doesn’t look like the full traceback, usually there is an error at the bottom. Would be great if you could post the whole thing.

[amsolu]

Thanks for your answer @erohmensing, let me give you more details on the sequence:

This ERROR message is after I went to the conversation tab, select the conversation (then the stories appear on the right pane) below the stories windows, there is a button “Copy to Interactive Learning” After clicking there, rasa x UI is waiting in interactive learning mode with no data on it and after 30 seconds or a minute, I have got this error message.

If the conversation is short, no problem so I believe it is related to the size of the conversation.

I also open another ticket that I would like to get your input, and it is related to the stories generated by rasa x and the ones generated in the interactive learning through the command line, it looks like they are not following the same format

Thank you

[erohmensing (Ella Rohm-Ensing)]

Weird, I don’t think it should be training a model for IL if you already have one trained (check the models page). Could you please paste all of the logs that happen from when you click the “Copy to Interactive Learning” button on a long story?