Reputation: 33
I'm trying to build a Dialogflow chatbot that fetches data from an external API via Google Cloud Functions, but without using Firebase. Despite extensive searching, I have not found any good examples or templates of this; it seems like all the examples available use Firebase functions.
I'm a newbie programmer and unfamiliar with Node.js, Promises and all that fancy stuff, but I have gathered that accessing an external API through Dialogflow should be possible even without Firebase (I'm using a paid version of Google Cloud).
I tried to create my Google Cloud Function using this Dialogflow weather API example, which is the closest thing I could find to what I need, even though this too uses Firebase: https://github.com/dialogflow/fulfillment-weather-nodejs/blob/master/functions/index.js#L72
The problem is that my code fails somewhere around the "res.on('end'..." line and I can't figure out why. The Google Cloud Stackdriver log only gives the rather uninformative message "Ignoring exception from a finished function", but does not tell me what the exception is.
Here is an edited version of my index.js code:
'use strict';
const rpn = require('request-promise-native');
const http = require('http');
const hostAPI = 'my host API URL goes here';
const url = require('url');
const {WebhookClient} = require('dialogflow-fulfillment');
exports.myGoogleCloudSearch = (req, res) => {
const agent = new WebhookClient({request: req, response: res}); // Dialogflow agent
// These are logged in Google Cloud Functions
console.log('Dialogflow Request headers: ' + JSON.stringify(req.headers));
console.log('Dialogflow Request body: ' + JSON.stringify(req.body));
// Default welcome intent, this comes through to Dialogflow
function welcome(agent) {
agent.add('This welcome message comes from Google Cloud Functions.');
}
// Default fallback intent, this also comes through
function fallback(agent) {
agent.add('This is the fallback response from Google Cloud Functions.');
}
function searchMyInfo(agent) {
// get parameters given by user in Dialogflow
const param1 = agent.parameters.param1;
const param2 = agent.parameters.param2;
const param3 = agent.parameters.param3
// this is logged
console.log('Parameters fetched from Dialogflow: ' + param1 + ', ' + param2 + ', ' + param3);
var myUrl = hostAPI + param1 + param2 + param3;
// the URL is correct and also logged
console.log('The URL is ' + myUrl);
// Everything up to here has happened between Dialogflow and Google Cloud Functions
// and inside GCF, and up to here it works
// Next, contact the host API to get the requested information via myUrl
// Using this as an example but *without* Firebase:
// https://github.com/dialogflow/fulfillment-weather-nodejs/blob/master/functions/index.js#L41
function getMyInfo(param1, param2, param3) {
console.log('Inside getMyInfo before Promise'); // this is logged
return new Promise((resolve, reject) => {
console.log('Inside getMyInfo after Promise'); // this is logged
console.log('Should get JSON from ' + myUrl);
rpn.get(myUrl, (res) => {
// The code is run at least up to here, since this is logged:
console.log('Inside rpn.get');
// But then the Google Cloud log just says
// "Ignoring exception from a finished function"
// and nothing below is logged (or run?)
let body = ''; // variable to store response chunks
res.on('data', (chunk) => {body += chunk;}); // store each response chunk
res.on('end', () => {
// this is not logged, so something must go wrong here
console.log('Inside res.on end block');
// Parse the JSON for desired data
var myArray = JSON.parse(body); // fetched JSON parsed into array
console.log(myArray); // not logged
// Here I have more parsing and filtering of the fetched JSON
// to obtain my desired data. This JS works fine for my host API and returns
// the correct data if I just run it in a separate html file,
// so I've left it out of this example because the problem seems
// to be with the Promise(?).
// Create the output from the parsed data
// to be passed on to the Dialogflow agent
let output = agent.add('Parsed data goes here');
console.log(output);
resolve(output); // resolve the promise
}); // res.on end block end
// In case of error
res.on('error', (error) => {
// this is not logged either
console.log('Error calling the host API');
reject();
}); // res.on error end
}); // rpn.get end
}); // Promise end
} // getMyInfo end
// call the host API: this does not seem to work since nothing is logged
// and no error message is returned
getMyInfo(param1, param2, param3).then((output) => {
console.log('getMyInfo call started');
// Return the results of the getMyInfo function to Dialogflow
res.json({'fulfillmentText': output});
}).catch(() => {
// no error message is given either
res.json({'fulfillmentText' : 'There was an error in getting the information'});
console.log('getMyInfo call failed');
});
} // searchMyInfo(agent) end
// Mapping functions to Dialogflow intents
let intentMap = new Map();
intentMap.set('Default Welcome Intent', welcome); // this works
intentMap.set('Default Fallback Intent', fallback); // this works
intentMap.set('my.search', searchMyInfo); // this does not work
agent.handleRequest(intentMap);
}; // exports end
So my question is: how can I make this code work to return the fulfillment response to Dialogflow? The default welcome and fallback responses do come through from Google Cloud Functions, but my custom intent webhook response does not (even though "Enable webhook call" is set in Dialogflow for my.search).
Upvotes: 2
Views: 5212
Reputation: 775
I did have the same issue and as you said I think is related to promises and asynchronous behavior of JavaScript. Because when you call a cloud function this executes and then response, but this function doesn't wait for external API call.
I tried request client, but when I saw the logs view, the external api response after cloud function response.
So I choose use axios (Promise based HTTP client for node.js), then the cloud function works.
This is a simple example for Dialogflow + external API + Google Cloud Functions:
index.js
'use strict';
const functions = require('firebase-functions');
const { WebhookClient } = require('dialogflow-fulfillment');
const { Card, Suggestion } = require('dialogflow-fulfillment');
var axios = require("axios");
process.env.DEBUG = 'dialogflow:debug'; // enables lib debugging statements
exports.dialogflowFirebaseFulfillment = functions.https.onRequest((request, response) => {
const agent = new WebhookClient({ request, response });
console.log('Dialogflow Request headers: ' + JSON.stringify(request.headers));
console.log('Dialogflow Request body: ' + JSON.stringify(request.body));
function welcome(agent) {
agent.add(`Welcome to my agent!`);
}
function fallback(agent) {
agent.add(`I didn't understand`);
agent.add(`I'm sorry, can you try again?`);
}
function helloWorld() {
return axios({
method: "GET",
url: "https://run.mocky.io/v3/197de163-acc3-4c86-a13f-79314fe9da04",
data: "",
})
.then((response) => {
console.log(response.data.body.greeting); //Hello World
agent.add(response.data.body.greeting);
})
.catch((error) => {
console.log(error);
});
}
let intentMap = new Map();
intentMap.set('Default Welcome Intent', welcome);
intentMap.set('Default Fallback Intent', fallback);
intentMap.set('greeting.helloWorld', helloWorld);
agent.handleRequest(intentMap);
});
Remember also add axios
package to package.json
:
{
"name": "dialogflowFirebaseFulfillment",
"description": "This is the default fulfillment for a Dialogflow agents using Cloud Functions for Firebase",
"version": "0.0.1",
"private": true,
"license": "Apache Version 2.0",
"author": "Google Inc.",
"engines": {
"node": "10"
},
"scripts": {
"start": "firebase serve --only functions:dialogflowFirebaseFulfillment",
"deploy": "firebase deploy --only functions:dialogflowFirebaseFulfillment"
},
"dependencies": {
"actions-on-google": "^2.2.0",
"firebase-admin": "^5.13.1",
"firebase-functions": "^2.0.2",
"dialogflow": "^0.6.0",
"dialogflow-fulfillment": "^0.5.0",
"axios": "^0.20.0"
}
}
Finally this is a post http query that I made, maybe you find it useful.
function consulta() {
// console.log(agent.parameters.consulta);
// console.log(agent.query);
var consulta = agent.query.replace(/(\r\n|\n|\r)/gm, " ");
return axios({
method: "POST",
url: "http://jena-fuseki-api:3030/Matricula",
headers: {
Accept: "application/sparql-results+json,*/*;q=0.9",
"Content-Type": "application/x-www-form-urlencoded",
},
params: {
query: consulta,
},
})
.then((response) => {
var elements = response.data.results.bindings;
for (var i = 0; i < elements.length; i++) {
var result = "";
var obj = elements[i];
var j = 0;
var size = Object.size(obj);
for (var key in obj) {
var attrName = key;
var attrValue = obj[key].value;
result += attrName + ": " + attrValue;
if (j < size - 1) result += " | ";
j++;
}
console.log(result);
agent.add(result);
}
console.log("----------------------------");
})
.catch((error) => {
console.log("Failed calling jena-fuseki API");
console.log(error);
});
}
Some images from Dialogflow:
Upvotes: 1
Reputation: 33
After a lot of trial and error, I came up with this index.js that works for my particular use case as far as I've been able to test it. I'm including it here in case someone else wants to try it out with a different API. If you do test it, please comment here! I would be interested in knowing how it works for another case.
/**
* Copyright 2017 Google Inc. All Rights Reserved.
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
'use strict';
// Include nodejs request-promise-native package as dependency
// because async API calls require the use of Promises
const rpn = require('request-promise-native');
const hostAPI = 'https://my.api.here/'; // root URL of the API
const { WebhookClient } = require('dialogflow-fulfillment');
exports.googleCloudSearch = (req, res) => {
const agent = new WebhookClient({ request: req, response: res }); // Dialogflow agent
console.log('Dialogflow Request headers: ' + JSON.stringify(req.headers)); // testing
console.log('Dialogflow Request body: ' + JSON.stringify(req.body)); // testing
// Default welcome intent
function welcome(agent) {
agent.add('Welcome to my chatbot!');
}
// Default fallback intent
function fallback(agent) {
agent.add('Sorry, I don\'t understand.');
}
// Default conversation end
function endConversation(agent) {
agent.add('Thank you and have a nice day!');
}
// Function for passing data to the myapi.search intent in Dialogflow
function searchMyApi(agent) {
return new Promise((resolve, reject) => {
// get parameters given by user in Dialogflow
const param1 = agent.parameters.param1;
const param2 = agent.parameters.param2;
// and so on...
console.log(`Parameters from Dialogflow: ${param1}, ${param2}`); // testing
// If necessary, format the parameters passed by Dialogflow to fit the API query string.
// Then construct the URL used to query the API.
var myUrl = `${hostAPI}?parameter_1=${param1}¶meter_2=${param2}`;
console.log('The URL is ' + myUrl); // testing
// Make the HTTP request with request-promise-native
// https://www.npmjs.com/package/request-promise
var options = {
uri: myUrl,
headers: {
'User-Agent': 'Request-Promise-Native'
},
json: true
};
// All handling of returned JSON data goes under .then and before .catch
rpn(options)
.then((json) => {
var result = ''; // the answer passed to Dialogflow goes here
// Make a string out of the returned JSON object
var myStringData = JSON.stringify(json);
console.log(`This data was returned: ${myStringData}`); // testing
// Make an array out of the stringified JSON
var myArray = JSON.parse(myStringData);
console.log(`This is my array: ${myArray}`); // testing
// Code for parsing myArray goes here, for example:
if (condition) {
// For example, the returned JSON does not contain the data the user wants
result = agent.add('Sorry, could not find any results.');
resolve(result); // Promise resolved
}
else {
// If the desired data is found:
var output = ''; // put the data here
result = agent.add(`Here are the results of your search: ${output}`);
resolve(result); // Promise resolved
}
}) // .then end
.catch(() => { // if .then fails
console.log('Promise rejected');
let rejectMessage = agent.add('Sorry, an error occurred.');
reject(rejectMessage); // Promise rejected
}); // .catch end
}); // Promise end
} // searchMyApi end
// Mapping functions to Dialogflow intents
let intentMap = new Map();
intentMap.set('Default Welcome Intent', welcome);
intentMap.set('Default Fallback Intent', fallback);
intentMap.set('End Conversation', endConversation);
intentMap.set('myapi.search', searchMyApi);
agent.handleRequest(intentMap);
}; // exports end
Upvotes: 1
Reputation: 50731
There may be other issues (I didn't read your code too carefully), but one is that although you're doing async operations, and returning a Promise in your call to getMyInfo()
, you also need to have the Intent Handler searchMyInfo()
return a Promise. This is so the handler dispatcher knows to wait for the promise to be completed before returning the response.
It also looks a little... odd... the way you're handling the response. Once you're using the dialogflow-fulfillment library, you should probably use it to generate the JSON (using agent.add()
), rather than trying to send the JSON yourself. I didn't test this, but it could be that trying to send the JSON yourself, and then having the library try to set the JSON, could result in invalid JSON that Dialogflow is rejecting.
Upvotes: 0