Options
All
  • Public
  • Public/Protected
  • All
Menu

Module dialogflow

DialogFlow Module

This module implement the IntentRecognizer interface with DialogFlow ES API.

Install

npm install @machinat/core @machinat/dialogflow
# or with yarn
yarn add @machinat/core @machinat/dialogflow

Docs

Check the Recognizing Intent document and the API references.

Setup

You can use this module in two different modes:

Delegated Mode

In this mode, the DialogFlow project is managed by the package. You maintain the training data along with the codes, and delegate the configuring procedures to the package.

First you need to create a GCP project and a service account to access the API. Follow this guide and set the GOOGLE_APPLICATION_CREDENTIALS environment variable when running your app.

Then add @machinat/dialogflow module like this:

import Machinat from '@machinat/core';
import DialogFlow from '@machinat/dialogflow';

const app = Machinat.createApp({
modules: [
DialogFlow.initModule({
projectId: '_YOUR_DIALOGFLOW_PROJECT_ID_',
recognitionData: {
defaultLanguage: 'en',
languages: ['en'],
intents: {
greeting: {
trainingPhrases: {
en: ['hello', 'hi']
}
}
},
},
}),
],
});

Finally you have to call DialogFlowRecognizer.train() method every time you update the intents data. Like:

// cli/updateDialogFlow.js
import Machinat from '@machinat/core';
import DialogFlow from '@machinat/dialogflow';

const app = Machinat.createApp({/* ... */});
app
.start()
.then(() => {
const [recognizer] = app.useServices([DialogFlow.Recognizer]);
return recognizer.train();
});

It's recommended to run this every time you deploy a new version of your app. It'll not retrain if the intents data remain the same, so it's really cheap to call.

Manual Mode

In this mode, you have to manage the DialogFlow project on your own. The package only works as a client for detecting intents.

First you have to prepare a ready-to-use DialogFlow agent. You can follow this guide to create one and add the intents in the DialogFlow console.

Next follow this section to create a service account and set the GOOGLE_APPLICATION_CREDENTIALS environment variable.

Then add @machinat/dialogflow module like this:

import Machinat from '@machinat/core';
import DialogFlow from '@machinat/dialogflow';

const app = Machinat.createApp({
modules: [
DialogFlow.initModule({
projectId: '_YOUR_DIALOGFLOW_PROJECT_ID_',
// prevent package to edit the project
manualMode: true,
// detect intent with the default agent in dev
useDefaultAgent: process.env.NODE_ENV !== 'production',
// specify the environment to be used on production
environment: 'production',
recognitionData: {
defaultLanguage: 'en',
languages: ['en'],
intents: {},
},
}),
],
});

If you call DialogFlowRecognizer.train() method under manual mode, it creates a snapshot version on the environment. You can do it every time you deploy a new version of app, so the DialogFlow agent would have a revertible version along with the version of app.

Index

Other

DetactIntentPayload: protos.google.cloud.dialogflow.v2.IQueryResult
DetactIntentResponse: protos.google.cloud.dialogflow.v2.IDetectIntentResponse
DialogflowConfigs<Language, Intent>: { agentName?: string; agentTimeZone?: string; clientOptions?: ClientOptions; environment?: string; manualMode?: boolean; projectId: string; recognitionData: RecognitionData<Language, Intent>; useDefaultAgent?: boolean }

Type parameters

  • Language: string

  • Intent: string

Type declaration

  • Optional agentName?: string

    The display name of the DialogFlow agent. Default to 'machinat-agent'

  • Optional agentTimeZone?: string

    The default time zone of the DialogFlow agent. Default to 'GMT'

  • Optional clientOptions?: ClientOptions

    The constructor options of a GCP client. Omit this if you already set the GOOGLE_APPLICATION_CREDENTIALS env. Check https://github.com/googleapis/gax-nodejs/blob/main/client-libraries.md#constructor-options

  • Optional environment?: string

    The environment to be used by the module. Default to 'machinat-entry'

  • Optional manualMode?: boolean

    Set to true to prevent the package from editing DialogFlow data, i.e. you have to mangage the DialogFlow project on your own

  • projectId: string

    The id of dialogflow project

  • recognitionData: RecognitionData<Language, Intent>

    The intents data for training.

  • Optional useDefaultAgent?: boolean

    If it's set to true, the default agent is used to detect intent instead under the environment.

SessionClient: v2.SessionsClient

Generated using TypeDoc