Actions on Google Assistant - a full stack guide

Slides for the talk at Google Devfest Mumbai 2018
While developing Actions for Google Assistant, developers often overlook the importance of testing, continuous integration and other software development best practices. This can lead to several problems down the line when developers try to scale their Actions or incorporate new features. Analytics is also a crucial part of an Action as it helps the developers gather more insight into the user’s behavior. This talk would help them understand the importance of these basic things and how to keep them in mind while building Actions for Google Assistant so that they may build truly robust, scalable and fault tolerant Actions.

Transcript

1. Alkesh Srivastava, Smalt and Beryl @_alkesh_ Mumbai Actions on Google - a full stack guide
2. Building Actions - a quick overview Mumbai
3. Design Develop Deploy Publish Process
4. Design 1. Create a persona 2. Draft a conversation 3. Understand context 4. Handle speech errors
5. Develop 1. Create a Dialogflow agent 2. Define entities and intents 3. Define user phrases 4. Define actions and responses 5. Connect webhook
6. Deploy 1. Register your Action 2. Test your Action 3. Review publishing checklist 4. Publish your Action
7. That’s all that’s needed to publish your Action! Mumbai
8. But I need my Action to scale.. Mumbai
9. ..and also incorporate software development best practices Mumbai
10. Backend using Typescript Mumbai
13. Why Typescript? 1. Optional static typing 2. Language features 3. Better suited for larger projects having multiple developers
15. C:\Users\alkesh\Desktop\Puffin> firebase init functions ######## #### ######## ######## ######## ### ###### ######## ## ## ## ## ## ## ## ## ## ## ## ###### ## ######## ###### ######## ######### ###### ###### ## ## ## ## ## ## ## ## ## ## ## ## #### ## ## ######## ######## ## ## ###### ######## You're about to initialize a Firebase project in this directory: C:\Users\alkesh\Desktop\Puffin ? Are you ready to proceed? Yes === Project Setup First, let's associate this project directory with a Firebase project. You can create multiple project aliases by running firebase use --add, but for now we'll just set up a default project. ? Select a default Firebase project for this directory: [create a new project] === Functions Setup A functions directory will be created in your project with a Node.js package pre-configured. Functions can be deployed with firebase deploy ? What language would you like to use to write Cloud Functions? TypeScript ? Do you want to use TSLint to catch probable bugs and enforce style? Yes + Wrote functions/package.json + Wrote functions/tslint.json + Wrote functions/tsconfig.json + Wrote functions/src/index.ts ? Do you want to install dependencies with npm now? Yes
16. Implementing continuous integration Mumbai
17. What is continuous integration? The practice of testing each change done to your codebase automatically and as early as possible
18. Why use CI? 1. Bugs are detected early 2. Rapid feedback on bugs 3. Less time spent on debugging 4. Goodbye long integration!
19. Workflow
20. How to implement CI in firebase 1. Create a .travis.yml 2. Generate the firebase token 3. Deploy to project
21. language : node_js node_js : - “8.0” before_script : - npm run setup - npm run build script : - npm run test install : - npm install -g firebase-tools after-success : - firebase deploy --token $FIREBASE_TOKEN
22. Testing Mumbai
24. Why unit testing? 1. Find bugs early 2. Facilitates changes 3. Simplifies integration
25. Automated testing in Actions on Google 1. Allows you to send automatic text queries 2. Receive SSML, text or even rich responses 3. Uses your account to send the queries
26. const { expect } = require('chai'); const { ActionsOnGoogleAva } = require('actions-on-google-testing'); const action = new ActionsOnGoogleAva(require('./test/test-credentials.json')); // First step is to give the test a name action.startTest('Order cupcake', action => { // Return a promise that starts a conversation with your test app return action.startConversationWith(“Cupcake Factory”) .then( res => { return action.send('I want order some cupcakes'); }) .then( res => { return action.send('Vanilla'); }) .then(({ textToSpeech }) => { expect(textToSpeech[0]).to.be.equal(‘Alright, how many do you want?’); }) return action.endTest(); });
27. A quick reference 1. action.start() - start conversation with your action 2. action.send() - send some phrase to your action 3. action.cancel() - end your conversation 4. action.locale() - set the locale for your conversation
28. Publishing your Action
30. Analytics Mumbai
32. Benefits of using Stackdriver 1. Examine error logs for fulfilment 2. Ability to filter logs 3. No custom code required
33. Want more insights from your data? Mumbai
34. Google BigQuery and Metabase are here to help Mumbai
35. What is Google BigQuery? 1. It is a cloud-based big data analytics web service for processing very large read-only data sets 2. It is designed for analyzing data on the order of billions of rows, using a SQL-like syntax.
36. async function reportInteraction(app: DialogflowApp, response: string) { const dataset = bigquery.dataset('lessons'); const errors = await dataset.table('interactions').insert([{ time: new Date(), user: app.getUser().userId, input: app.getRawInput(), response: response, inputType: inputType(app.getInputType()), }]); if (errors && errors.length && errors[0].insertErrors) { console.error('Bigquery Insert failed', errors[0].insertErrors); } }
37. And Metabase? It’s a business intelligence tool for helping you gather insights from your data and ask questions from it
38. Setting up metabase with BigQuery 1. Connect to BigQuery database 2. Generate and copy client id to metabase 3. Copy and paste Auth Code to metabase 4. Start fetching data using SQL
39. The custom metabase dashboard
40. Alkesh Srivastava, Smalt and Beryl @_alkesh_ Mumbai Thank you!