Rasa x does not give response to detected intent

[Deprecated] Rasa X Community Edition
1

Hey guys, I’m new to rasa and have been playing around with it for a week. I’ve been looking around and have found some topics but they weren’t exactly the same issue I have. For the sake of simplicity I have deployed rasa x to a test cluster so you can try it and ask for password if you want to debug it. I’m using the policy that is suggested in https://rasa.com/docs/rasa/nlu/choosing-a-pipeline/ and wanted to simply use mapping policy to start with something simple (faq bot). Initially I setup the nlu with persian language but then I thought maybe the failure is because of the language but the same thing happen with engligh. You can see I have added 2 intents (questions) and the corresponding responses and triggers in domain for mapping policy. When I try to do a conversation with the bot I can see in the api call it detects the intent but then nothing happens. you can try the bot here: http://rasa.genyus.tech/guest/conversations/production/bc070a83d8734d9f93befc81fe9dcaaf With this question: who created the world? You can see all the config in this gist: Rasa x config · GitHub

Thanks in advance and sorry if it’s a dumb question :smiley:

2

Hey @moein, welcome to the forum!

I think that predicted response doesn’t reach your Rasa X instance. That can happen when for some reason you have issues with DB or RabbitMQ image.

I’d suggest you to start with getting logs from “db” and “rabbit” apps, if you deployed RasaX with docker-compose you can do that with:

docker-copmose logs db
docker-copmose logs rabbit

Check if they have any ERROR entries.

P.S. for the FAQ bot you can you can use response selector, that will simplify your training data.

3

Hi,

Thanks for the fast response. Also really good to know about response selector (I would say it’s a killer feature but not really visible until you actually reach the section about it).

I checked the logs for rabbitmq and postgresql and there is no error logs.

But I did find this error in rasa production pod(I’m using the helm chart):

2020-04-06 09:21:12 ERROR rasa.core.processor - Encountered an exception while running action ‘first_answer’. Bot will continue, but the actions events are lost. Please check the logs of your action server for more information.

Which one is the action server?

NAME                                    READY   STATUS    RESTARTS   AGE
rasa-app-8688844469-skgh8               1/1     Running   0          2d20h
rasa-duckling-55bbfc5754-lqrhn          1/1     Running   0          2d20h
rasa-event-service-7bd75db4c5-qlffj     1/1     Running   3          2d20h
rasa-nginx-998bcbdf4-nktvm              1/1     Running   0          2d20h
rasa-postgresql-0                       1/1     Running   0          2d20h
rasa-rabbit-0                           1/1     Running   0          2d20h
rasa-rasa-production-54b875dccf-9rv29   1/1     Running   2          2d20h
rasa-rasa-worker-7c9777b4d8-svkpv       1/1     Running   2          2d20h
rasa-rasa-x-55766b8944-j5v85            1/1     Running   1          2d20h
rasa-redis-master-0                     1/1     Running   0          2d20h
4

hey @moein,

it’s rasa-app-8688844469-skgh8 . Looks like it cannot run your first_answer.

please tag me if I can still help you!

5

Hey @degiz,

So I decided to actually try the ResponseSelector and not actually test it on rasa-x but rather from rasa itself.

Here you have the repo: GitHub - moein/rasa_test

I got the data from another forum question just to test this feature(responseselector) but still no luck.

This is what I’m running on the repository that I mentioned:

docker run -it -v `pwd`:/app --entrypoint=/bin/bash rasa/rasa:1.9.5
rasa train
rasa shell --debug

And after writing who are you in the input it gives this debug message:

2020-04-08 21:50:58 DEBUG rasa.core.tracker_store - Creating a new tracker for id ‘378522bcb23f45d48f901ffbc8768486’. 2020-04-08 21:50:58 DEBUG rasa.core.processor - Starting a new session for conversation ID ‘378522bcb23f45d48f901ffbc8768486’. 2020-04-08 21:50:58 DEBUG rasa.core.processor - Action ‘action_session_start’ ended with events ‘[<rasa.core.events.SessionStarted object at 0x7fbff02efda0>, <rasa.core.events.ActionExecuted object at 0x7fbff02efd68>]’. 2020-04-08 21:50:58 DEBUG rasa.nlu.selectors.response_selector - Adding following selector key to message property: default 2020-04-08 21:50:58 DEBUG rasa.core.processor - Received user message ‘who are you’ with intent ‘{‘name’: None, ‘confidence’: 0.0}’ and entities ‘’ 2020-04-08 21:50:58 DEBUG rasa.core.processor - Logged UserUtterance - tracker now has 4 events. 2020-04-08 21:50:58 DEBUG rasa.core.policies.ensemble - Predicted next action using policy_0_MappingPolicy 2020-04-08 21:50:58 DEBUG rasa.core.processor - Predicted next action ‘action_listen’ with confidence 0.00. 2020-04-08 21:50:58 DEBUG rasa.core.processor - Action ‘action_listen’ ended with events ‘’. 2020-04-08 21:50:58 DEBUG rasa.core.lock_store - Deleted lock for conversation ‘378522bcb23f45d48f901ffbc8768486’.

I really like the architecture of rasa and how powerful it is but I’m not sure if it’s me or actually it’s hard to get rasa working :frowning:

6

hey @moein

I’ve found several issues:

  1. Your config.yml is missing a section with a classifier:
- name: DIETClassifier
  epochs: 100

ResponseSelector will not work without it. Also, it’s by default in the config.yml when you init rasa project with rasa init.

  1. In order for ResponseSelector to work you need to add at a least one more intent that is not faq, standard greeting would do.

With the mentioned changes I was able to train and talk to the bot from your repo.

I hope that helps!

1 Like
7

Amazing!

Thanks a lot @degiz Finally got it working!

I will make sure to share our final project with you guys as a showcase for non english project :slight_smile:

Cheers

1 Like
8

can we configure rasa, if a person types Hi name, @name, Hello name rasa should not respond.