Google Actions CLI 3.1.0 version and actions.intent.TEXT

I want to be able to talk with Google Assistant, but connect the Actions project directly to an NLP service I already have running on my server. In other words, NOT use dialogflow.

All the following examples show how to do this.

With Rasa
https://blog.rasa.com/going-beyond-hey-google-building-a-rasa-powered-google-assistant/

With LUIS
https://www.grokkingandroid.com/using-the-actions-sdk/
https://dzone.com/articles/using-the-actions-sdk-for-google-assistant-develop

With Watson
https://www.youtube.com/watch?v=no0R0bSkHXc

They use the actions.intent.MAIN as the invocation and actions.intent.TEXT for all other utterances from the talker.

This is what I need. I don’t want to create a load of intents, with utterance phrases, inside the Action because I just want all the phrases spoken by the talker to be passed to my server, and for my NLP service to deal with them.

So I set up a new Action project, install the Actions CLI and then spend 3 days trying all possible combinations without success, because all these examples are using gactions cli 2.1.3 and Google have now moved on to gactions cli 3.1.0.

Not only have the commands changed, but so too has the file formats and structure.

It appears there is also a new Google Actions Console, and actions.intent.TEXT is no longer available.

My Action is webhook connected to my server, but I cannot figure out how to get the action.intent.TEXT included and working.

Everything I find, even here

Publishing Actions on google without Dialogflow is pre version update and follows the same pattern.

Can anyone point to an up-to-date, v3.1.0, discussion, tutorial or example about how to send all talker phrases through to an NLP that isn’t dialogflow, or has Google closed that avenue?

Is it possible to somehow go back and use the 2.1 CLI either with the new Console or revert the console back. (I have both CLI versions, I can see how different their commands are)

Is it possible to go back and use 2.1?

There is no way to go back to AoG 2. You probably also don't want to do so - newer features aren't available with v2 and are only available with v3.

Can I use my own NLP with v3?

Yes, although it isn't as obvious, and there are some changes in semantics.

As an overview, what you'll need to do is:

• Create a Type that can accept "Free form text". I usually call this type "Any".

  In the console, it looks something like this:

  

• Create a Custom Intent that has a single parameter of this Any Type and at least one phrase that captures everything for this parameter. (So you should add one training phrase, highlight the entire phrase, and set it for the parameter. Sometimes I also add additional phrases that includes words that I don't want to capture.) I usually call the Intent "matchAny" and the parameter "any".

  In the console, it could be something like this:

  

• Finally, you'll have a Scene that you transition to from the Main invocation. When it matches the "matchAny" Intent, it should call your webhook with a handler name. Your webhook will be called with the "any" parameter set with the user utterance. (Note that the JSON has also changed.

  Again, the console might have it looking something like this:

  

That seems like a lot of work. Isn't there just some way to do all that from the command line?

Yes. You can do all of that in the configuration files that the CLI accesses and then upload it. (You can then also use the console to review the configuration, if necessary, to make sure they're configured as you expect. You can shift back and forth between them as appropriate.)

Google also has a github repository that contains most of the files pre-configured for this sort of setup.

You will need to update the configuration from the repository to handle the webhook correctly (it includes code to illustrate what is happening using the inline code editor) and to add your project ID.