Internet Archive Google Action

Setup Instructions

See the developer guide and release notes at https://developers.google.com/actions/ for more details.

Steps for testing with Google Assistant

Create and setup project in Actions Console

1. Use the Actions on Google Console to add a new project with a name of your choosing and click Create Project.
2. Click skip
3. Go to actions under build, and click Add your first action
4. select custom intent, then build

   Create Action for the project

5. Click CREATE (continued from above step)
6. Go to settings and click export and import, then click restore from zip.
7. zip the contents of models/dialogflow/ and use that zip to restore the project.

   Run Local Server

   Run server local with colorful logs

8. npm install -g firebase-tools
9. touch functions/.runtimeconfig.json and add fake keys (temporary workaround) refer this
10. firebase init (remove .firebaserc first)
11. firebase use --add (and choose your own firebase project)
12. cd functions and npm install
13. DEBUG=ia:* npm start

    Expose local server

    To expose server to google assistant use ngrok its free plan should be enough.

To publish 5000 port use:

ngrok http 5000

you should find url https://<id>.ngrok.io/.

Connect webhook to dialogflow

Go to the fullfillment section of your dialogflow draft copy of our app and after that you should use this url:

https://<id>.ngrok.io/<your project name>/us-central1/assistant

Steps for testing with Alexa

The Alexa app workflow for dev/production is the following:

• We create a zip file with the app code.
• We upload the zip to an AWS S3 bucket.
• We edit the AWS Lambda app to load the new app zip file.
• In case the AWS Lambda app changes a lot (e.g. we upgrade Node version), its Lambda ID will change. In that case, we need to edit the Alexa skill in https://developer.amazon.com/alexa/console/ask# to load the new Lambda app. This final step is not required most of the times.
• We use https://developer.amazon.com/alexa/console/ask# to test the app.

More details follow.

Node 12 is required. There is a problem with newer Node versions, bespoken-tools which is a key dependency, is not supporting it. Use nvm install 12 and nvm use 12 to switch to Node 12 before running any other command.

The general idea is that there are 2 AWS S3 buckets, dev and production. You can access them via https://aws.amazon.com/ -> sign in -> S3. You need to create a zip with the app, you upload it to the relevant S3 bucket and then you go to AWS Lambda dev / production to select which app should be loaded via its S3 URL. (AWS -> Lambda -> edit app -> "upload from S3 .zip file").

S3 buckets have multiple app versions. This way, you can also easily rollback to a previous version if needed.

The script that produces the app.zip that we upload to S3 is bin/deploy-to-alexa-skill.sh. An important AWS Lambda limitation is that the unzip code filesize should be < 250MB. To achieve that, we do:

(cd ./deploy/alexa/; npm install --production; rm -rf node_modules/aws-sdk; node-prune node_modules)

aws-sdk is already preinstalled in AWS Lambda, so we don't need it.

npm install -g node-prune is a pre-requisite. This tool removes files that are useless in production and reduces the total filesize a lot.

After deploying to dev or production, use https://developer.amazon.com/alexa/console/ask# ACTIONS -> Test for testing.

Setup Env

Options

• profile performance of requests Env Variable: PROFILE_REQUESTS=true (false by default) firebase functions:config:set performance.requests=true

How to make contributions?

:mag: get one good first issue assign yourself (if you have access) or write comment that you'd like to work on this issue. That's help to prevent work overlapping.

create git branch feature/<name-of-feature>, more

:computer: working on it

Use Mocha for continuous checking of your code quality and cover functionality by tests

npm run mocha -- --watch

:coffee: Complete checking of code by run unit tests and code style checking

npm test

:star2: We follow standard javascript code style.

Automatic style fixing, it doesn't solve all problems but could be very helpful

npm run lint -- --fix

:tada: Finally make Pull Request and give complete description what have you done and link the addressed issue.

Also it could be good practice to create your Pull Request earlier, but add WIP: at the beginning of its name! This way other developers could see what are you working right now.