Can you pip show rasa for me?
I feel like you might have been doing what I was, and if you visit this post Facebook Setup - Webhook not resolving - #6 by ManyChat - then youâll see how I dealt with a similar situation.
I am also assuming you have credentials.yml already filled out too. After that, youâll need to fix the endpoints.yml file and uncomment the webhook endpoint. Then run rasa with the following command:
rasa run -m models --credentials credentials.yml --enable-api --log-file out.log --endpoints endpoints.yml
Tried following your method but still the same error. Also can you tell me is it necessary to enable SSL certificate for setting up the webhook and if so which of the ones available online should I purchase?
@erohmensing any thoughts on what might be going wrong?
Hey @RenuDighe, ssl shouldnât be necessary just to set it up.
I think things are getting confused with your versions â you have rasa installed but also rasa_core and rasa_nlu. It looks like youâre still using the old versions because you were using python -m rasa-core.run .... which is for the old rasa_core.
I think the best thing is to get your bot completely migrated to rasa and then uninstall rasa_core and rasa_nlu.
I usually run the bot using ânpm startâ
Not sure how this became a thing, but (once you are migrated) you should be starting the bot with a rasa run command
Ok, I will try that!
I unistalled rasas nlu and core and had the chatbot train on rasa, now when I run it I get an error in the browser saying ârequested url not foundâ
That should be fine, thereâs no UI for that. Can you try the ngrok stuff and connecting to facebook again? If you use rasa shell your bot works, right?
Hi, it worked! Will you just confirm from the screenshots below if everything has been done correctly or not?
Apparently Facebook is reviewing the chatbot for which they sent me a legal contract as well. Is this all the normal procedure?
Iâm not sure about the legal contract stuff â Iâve set up a facebook chatbot in test mode but have never actually deployed one. I do remember it saying something about needing permission though.
the Ignoring message as there is no agent to handle it message doesnât look good though. How did you train your model and what does your config look like?
I trained it using ârasa train --data data/nlu_data.md --out models --verbose nluâ and then ârasa run --enable-api --log-file out.log --endpoints endpoints.yml --credentials credentials.ymlâ.
Config file -
language: en
pipeline: supervised_embeddings
policies:
- name: KerasPolicy epochs: 200 batch_size: 50 max_training_samples: 300
- name: FallbackPolicy
- name: MemoizationPolicy
- name: MappingPolicy
Ah, so youâre only training an NLU model there I believe (youâre only passing nlu data and, althought you put a bunch of args in between, are still running rasa train nlu. So there will be no core/dialogue management â is this intentional?
I assume not â if not you want to make sure your stories are in your data folder and then train with rasa train --data data/ --out models --verbose (which by the way is the same as just rasa train --verbose)
Looks like you have some cleaning up to do â are the messages unclear?
I donât think so. It is running well on the server. What do you mean by cleaning up? I can always try doing so
I mean you have to read the error messages and do what they say â i.e. add FormPolicy to your policy ensemble and add utterances for utter_time, utter_name, etc
You can also look with rasa data validate to get even more info
So I resolved the errors and when I train and run, I get this
If I go to the localhost in browser I see this
Is it working fine now? However the bot is not running on the server, can you tell me why?
Again just as a heads up, it works if I say ânpm startâ
I suppose you must have a completely different setup than normal because we donât use npm at all. If it works, it works though, I guess
