Rasa Livecoding: Create a new assistant in Rasa 2.0

Поделиться
HTML-код
  • Опубликовано: 12 дек 2024

Комментарии • 41

  • @nikotime
    @nikotime 4 года назад +4

    This is a brilliantly helpful and informative video, thanks!

  • @doyourealise
    @doyourealise 3 года назад +3

    WOw, i am trying to create a chatbot for mental health problems, amazing found this video. Love it/:)

  • @kaushal7able
    @kaushal7able 2 года назад

    Thanks, I got a idea how to get started with rasa. By the way your smile is very cute

  • @rajasaroj7818
    @rajasaroj7818 3 года назад +4

    I think you forgot to save the nlu.yml file in vscode, may be that's the reason it wasn't able to pickup the intent and examples while training :)

  • @SATSifaction
    @SATSifaction 3 года назад

    when deploying this in prod, where are users questions stored? if the bot doesnt answer it correctly, how do you get a list of the questions answered so you can analyze the top questions?

    • @RasaHQ
      @RasaHQ  3 года назад +1

      It's stored in the tracker store & for the analysis, Rasa X is a tool we created just for that: rasa.com/docs/rasa-x/

  • @harlenmineiro
    @harlenmineiro 4 года назад +2

    Hi, can I found source code of this video somewhere?

    • @RasaHQ
      @RasaHQ  3 года назад

      Yep! github.com/rctatman/personal_website_bot

  • @movierow5225
    @movierow5225 2 года назад

    Connect the rasa flask or django

  • @bhuvaneshkumarsrivastava906
    @bhuvaneshkumarsrivastava906 3 года назад

    I have one question:
    When a user enters a query then RASA uses that query & finds which of the available intent will have maximum confidence score (in range 0 to 1) and the intent with maximum confidence score will be selected by the model.
    I have a use case where this maximum score is very less (i.e RASA model got the maximum score intent for a user query but that score is not good enough to label a query with that intent ). In this case, does RASA provide a Thresholding feature for Intent score, below which it will just print a message like: "Sorry I was unable to understand what you have said. Please re-try." ?

    • @RasaHQ
      @RasaHQ  3 года назад

      Yep! You're looking for the fallback policy: rasa.com/docs/rasa/fallback-handoff/

  • @rajeshpolaki9157
    @rajeshpolaki9157 4 года назад

    nice lecture
    can you help me in adding buttons in payload while bot give options in button selection rather than giving text format

    • @RasaHQ
      @RasaHQ  3 года назад +1

      There's a known bug (github.com/RasaHQ/rasa/issues/7344) & the fix is waiting on tests (github.com/RasaHQ/rasa/pull/7373) but if you're willing to take the chance of it breaking something you can try adding it to your local version of Rasa.

  • @luisadrian353
    @luisadrian353 3 года назад

    Where is the second part of this video?

    • @RasaHQ
      @RasaHQ  3 года назад +1

      all the livecoding videos in order are in this playlist: ruclips.net/p/PL75e0qA87dlFvomavOFBIX6S9IvtGAXjM -Rachael

  • @karunamayiholisticinc
    @karunamayiholisticinc 3 года назад

    Hi, this is good. Do we not need to give Formpolicy in config.yml when using form? Also, what is the way to connect RASA chatbot to a live agent if the bot can not answer the question?

    • @RasaHQ
      @RasaHQ  3 года назад +1

      1. Not as of 2.0! Now all deterministic dialog policies (except for memoization) are part of the rule policy
      2. We have an example of a similar behaviour (passing from one bot to another) in this starter pack: github.com/RasaHQ/helpdesk-assistant
      - R

    • @karunamayiholisticinc
      @karunamayiholisticinc 3 года назад

      @@RasaHQ Thanks I figured that out there is difference between rasa 1.0 and 2.0.
      hi have another question. RASA has not been able to predict the proper response for some of my intents. When I run rasa shell --debug
      it is able to detect the intent with very high level of confidence level. But when it comes to predicting the response it shows this message:
      rasa.core.policies.memoization - There is no memorised next action
      And then right after that the TED predicts some other response which is not the correct one. I am not sure why it is doing that.
      Would you know anything in regards to this? Will be greatly helpful. Thanks

  • @norak7591
    @norak7591 3 года назад

    Nice video!
    I have couple of questions:
    1. How can I make two actions for a specific intent in a story. For example if I have the intent "thanks" I want the chatbot to first do the action "utter_you_are_welcome" then the action " utter_goodbye"
    2. to start my rasa project I used "rasa init" and the basic chat bot I start with has no action list in the domain.yml file. What is this list for and can the bot work without it?
    Thank you!

    • @RasaHQ
      @RasaHQ  3 года назад +1

      (Vincent here)
      1. You can have two actions that follow each other in your stories.yml file. If you go to the stories.yml from the 'rasa init'-bot you should see an example.
      2. The starter assistant doesn't come with any custom actions defined. We do add responses, but these are different from what we call "custom actions" because in Rasa custom actions run custom python code.

  • @hiteshsingh9859
    @hiteshsingh9859 3 года назад

    hie am working on F and Q chatbot .and my intent around 250 ..my accuracy is very less .can you please suggest which pipelines is best and number of epoch. can you please suggest what i write in config.yml ?

    • @RasaHQ
      @RasaHQ  3 года назад

      You're probably going to see the best results by reducing your number of intents. Some of them may be able to be a single intent w/ multiple entities instead. I have a live coding session where I reduce the # of intents for a bot I'm working on that might be helpful: ruclips.net/video/U0nExRfH2Bo/видео.html -- R

    • @hiteshsingh9859
      @hiteshsingh9859 3 года назад

      @@RasaHQ great !!i will watch and try it .thank you so much

  • @o5051
    @o5051 3 года назад

    Hello can anyone tell me if user login and chat with bot and user logout and again same user come in 2-3 hours then the chat is there not start from beginning.

  • @SagarKar10
    @SagarKar10 3 года назад

    %YAML 1.2 --- YAML: YAML Ain't Markup Language™ What It Is: YAML is a human-friendly data serialization language for all programming languages.

  • @saimadhuchandhana4718
    @saimadhuchandhana4718 3 года назад

    Hey, its a very informative video. I have a question: Can we have Rasa 1.10.17 in one conda environment and Rasa 2.0 in another conda environment in same windows 10 system? Please answer my question, it helps me a lot. Thanks in advance.

    • @RasaHQ
      @RasaHQ  3 года назад

      Yep, the conda environments will be seperate. (I would just be careful about your Python versions, since versions below 2.x aren't compatible with Python 3.8) - Rachael

    • @saimadhuchandhana4718
      @saimadhuchandhana4718 3 года назад

      @@RasaHQ Thank you for the response. I actually tried it. But Rasa is being installed outside the env. I have installed 1.10.17 in 1 env and then installed 2.0 in other env. The 1.10.17 is being overwritten by 2.0 version. Please help me with this.

    • @RasaHQ
      @RasaHQ  3 года назад +1

      @@saimadhuchandhana4718 it could be that you're accidentally installing Rasa system wide rather than in an active environment (super easy to do by mistake). This blog has a nice discussion: towardsdatascience.com/a-guide-to-conda-environments-bc6180fc533#5193

    • @saimadhuchandhana4718
      @saimadhuchandhana4718 3 года назад

      @@RasaHQ I was using same python version for both environments, used different versions and I could install it perfectly. Thanks a lot for the help.

  • @ppujari
    @ppujari 2 года назад

    Does anyone here know free software to publish Rasa newsletter?

  • @TheDarkusx
    @TheDarkusx 4 года назад

    Hi, can we be able to use rasa 2.0 with rasa x

    • @RasaHQ
      @RasaHQ  4 года назад

      Yep! You just need to make sure your Rasa X version is 0.33 or above. rasa.com/docs/rasa-x/changelog/rasa-x-changelog

  • @merakiclt5806
    @merakiclt5806 2 года назад

    is that rasa is real?