Cloud Function for ChicagoRoboto talk.

function.js

'use strict';

process.env.DEBUG = 'actions-on-google:*';
const Assistant = require('actions-on-google').ApiAiAssistant;

exports.myAction = (req, res) => {
  const assistant = new Assistant({request: req, response: res});
  console.log('Request headers: ' + JSON.stringify(req.headers));
  console.log('Request body: ' + JSON.stringify(req.body));

  // Fulfill action business logic
  function responseHandler (assistant) {
    // Complete your fulfillment logic and send a response
    assistant.ask('Im sorry to hear that your ' + assistant.getArgument('body-parts') + ' hurts. Does anything else hurt?', ['I didnt hear you', 'What else is causing you pain', 'I guess this concludes our consultation']);
  }

  const actionMap = new Map();
  actionMap.set('describe-symptoms', responseHandler);

  assistant.handleRequest(actionMap);
};

In the javascript code above, the mapping on line 18 is where we map the intent name as setup in api.ai to the response handler.

The responseHandler can query the parameters and context from the assistant class, as you can see on line 14.

As mentioned in my talk at the end, you respond back to api.ai by calling assistant.ask (if you want to leave the mic open), or assistant.tell (if you want to close the mic and the conversation). The extra parameters you see at the end of the ask method call are some prompts that Google Assistant will use if the user doesn't respond in a timely fashion.

Learn more about the ApiAiAssistant class here https://developers.google.com/actions/reference/ApiAiAssistant

The webhook request/response format is described here https://developers.google.com/actions/reference/webhook-format