Rasa init deletes playground - Hello World Fail (for a newbie)


As I newbie, I’ve been following Rasa Playground

I went through the steps on the tabs there (i.e. NLU Data, Responses, Stories, Forms, Rules). For each of those tabs, I read everything and did not edit anything. It’s clearly a newsletter subscription bot. Fantastic!

Next I went down to where it says…

Beginning of Webpage Excerpt

You have built your assistant! What’s next?#

You can download this project and continue building to create a more advanced assistant.

Download project

End of Webpage Excerpt

I clicked the “download project” button. A .zip representing my project got downloaded. Hooray!

I unzipped the project.zip in a directory, let’s call it hello_world1

I cd’d into hello_world1

Next, I called “rasa init” and hit the defaults for everything (e.g. probably I hit “Y” several times in succession).

Next thing I know I am talking to a bot. Terrific! However, the bot is clearly not the one that was designed in the tabs I mentioned above. It was more of a mood bot, described elsewhere. What happened to the newsletter subscribe bot?!?!

Further, I note that calling rasa init causes it to change what had been downloaded. Specifically, the config, nlu, stories, rules, domain yml files have all been replaced by something the “rasa init”

So, my question boils down to this: How can I simply initialize, start up, fire up or run rasa so that it does not blow away what I have written? I just want to start up rasa and have it load the files I have right there in front of it and not download the mood bot!

My guess is that there is some argument that I can pass to rasa init that causes it to read the files I’ve already put there, rather then blowing away what is already there and replacing all the .yml files with the moodbot.

Or maybe there is some other command than rasa init that I should be calling?

I must be missing something obvious but I have searched far and wide and found no guidance.

Thank You!


Sigh! I think I figured it out.

The answer is simple.

cd into the directory where I had my stuff (i.e. hello_world1)

then call…

unix% rasa train

…then call…

unix% rasa shell

The newsletter subscribe bot started and now I have finally succeeded with hello world.

Can someone make sure that

unix% rasa train


unix% rasa shell

gets added right below where it says “You can download this project and continue building to create a more advanced assistant.” at Rasa Playground pretty please?

It took like an hour or more of endless fruitless searching and then 30 minutes to write the note I just wrote. Then I had the inspiration to just type “rasa” at my unix command line. Then I figured out that rasa train followed by rasa shell was the way to go.


Can someone improve the documentation? Or point out where it was lying there plain as day and I missed it?

T h a n k Y o u !


rasa init deletes all in the folder and sets up the default moodbit.

You should have done rasa train to train the existing model :slight_smile:

1 Like

@ChrisRahme thanks for confirming that.

Presumably the next step after “rasa train” is “rasa shell”?

Chris or anyone, would it be worthwhile if I asked the powers that be if they could clarify this aspect of the documentation (as detailed in my first post)?

If so, how would I make that appeal?

Yup! Or rasa x (you have to install it first though).

In my opinion, the documentation should be clear, you have the right to ask for improvements. I also had some issues understanding the process of deployment on a server the first time I read the docs.

I think that requests for updating the documentation are made via GitHub issues.

1 Like

+1 to Chris’s reocmmendation, @Tester01, a bug would be appreciated. :slight_smile: :pray:


@rctatman thanks for the +1 I finally submitted the “bug” (technically a feature improvement - whatever). You can find it here: https://github.com/RasaHQ/rasa/issues/9943



@ChrisRahme thanks for the pointers and the bug report suggestion. As I just stated above to rctatman I finally submitted the bug report here: https://github.com/RasaHQ/rasa/issues/9943