DEV Community: Samuel

My 7 steps to build a chatbot efficiently

Samuel — Thu, 05 Aug 2021 10:05:22 +0000

Step 1. Identify the problem

What is the pain point you want to solve? For instance, you want to build a chatbot to help your team organize the sales meetings.

Step 2. Define the scope

What is the main functionality? Your chatbot should be able to schedule meetings. The bot will connect the sales representatives and the manager. When the meeting time is scheduled, the bot will send a reminder to all involved parties.

Step 3. Identify the endpoints

What is the content you want to share? What is the format? You need to think about the information that will be shared. For instance, the bot will have to send the meeting invite via email, to create a calendar event, and to send reminders.

Step 4. Design the conversation flow

What are the possible paths of the conversation? For example, to schedule a meeting, the manager can ask to invite another sales representative.

Step 5. Code the chatbot

Write the conversational script and use tools to code chatbot:

Some examples of tools:

a) IBM Watson Bot Builder

b) Wit.ai

c) Dialogflow

d) Botfuel

Step 6. Test the chatbot

Test and refine the chatbot.

Step 7. Deploy the chatbot

DialogFlow Chatbot with NewBot Framework

Samuel — Tue, 20 Aug 2019 14:52:32 +0000

The goal: to create a chatbot quickly in Javascript with the understanding of the natural language (DialogFlow).

Prerequisites

Have a Google Cloud and DialogFlow account
Have NodeJS and NPM installed on your machine
Knowledge of Javascript

Why NewBot Framework?

Universal, a single code to work everywhere
Creation of conversational website
Modular and structured
Using the NLP (Natural Language Processing)
Conversational script syntax to write user-chatbot conversations easily and quickly
Internationalization
Unit tests
Emulator with NewBot CLI

Compatibility

All platforms, i.e.

Facebook Messenger
Google Assistant
Amazon Alexa
Slack
Telegram
Viber
Skype
etc.

Above all

Install NewBot CLI to take advantage of command lines

npm install -g newbot-cli

Create a project

newbot new <your directory name>

Go to the generated folder

You can test the chatbot with the command newbot serve. See below

Retrieve the credentials of DialogFlow

Have a Google Cloud and DialogFlow account

Click on the service in Google Cloud

Create new Key

Download the private key and add it to the root of the project

Add DialogFlow in our project

Install the newbot-dialogflow module :

npm install newbot-dialogflow

In the main.js file, add the DialogFlow skill :

import dialogflowSkill from 'newbot-dialogflow'
import code from './main.converse'

export default {
    code,
    skills: {
        dialogflow: dialogflowSkill({
            projectId: 'newbot-fttkoh',
            credentials: 'newbot-fttkoh-69d17227a8b7.json'
        })
    }
}

projectId : The project ID (in Chatbot Settings)
credentials: Path to credentials file

Input unknown

Add this code to the main.converse file :

@Intent('input.unknown')
unknown() {
    > { :intent.response }
}

An input.unknow intention is already present in DialogFlow, it is triggered when no other intention is triggered
We then display the response received from DialogFlow

:intent.date.value is a magic variable that retrieves the date entered by the user

The language used is specific to NewBot. It is called ConverseScript and allows you to create conversational scripts with a syntax close to Javascript
More : https://newbot.io/docs/syntax/variables.html

Add our own intention: we buy a pizza!

Create a new intent in DialogFlow
Put a name ("Buy Pizza" by example)
Add trainings phrases
Add action name. Here, this name is input.buy
You can add parameters. Here, we recover the date in the sentence
Add response(s)

main.converse content contains a new intention!

@Intent('input.buy')
buy() {
    date = :intent.date.value
    > { :intent.response }
    callApi(date)
}


@Intent('input.unknown')
unknown() {
    > { :intent.response }
}

We retrieve the date entity and call a function that can trigger an API

callApi() is a function created by us, present in Javascript :

main.js

import dialogflowSkill from 'newbot-dialogflow'
import code from './main.converse'

export default {
    code,
    skills: {
        dialogflow: dialogflowSkill({
            projectId: 'newbot-fttkoh',
            credentials: 'newbot-fttkoh-69d17227a8b7.json'
        })
    },
    functions: {
        callApi(date) {
            console.log(`Call your own API. date is ${date}`)
        }
    }
}

Test in Emulator

Type the following command at the root of your project

newbot serve

Go to localhost:3000 and test the project in the emulator

Documentation

Here : https://newbot.io/en/docs

Create a universal chatbot with JavaScript and NewBot Framework

Samuel — Mon, 19 Aug 2019 13:20:27 +0000

We will create a modular and fast chatbot. We take advantage of NLP without an external platform (like DialogFlow)

Prerequisites

Have NodeJS and NPM installed on your machine
Knowledge of Javascript

Why NewBot Framework?

Universal, a single code to work everywhere
Creation of conversational website
Modular and structured
Using the NLP (Natural Language Processing)
Conversational script syntax to write user-chatbot conversations easily and quickly
Internationalization
Unit tests
Emulator with NewBot CLI

Compatibility

All platforms, i.e.

Facebook Messenger
Google Assistant
Amazon Alexa
Slack
Telegram
Viber
Skype
etc.

and directly in your browser (offline)

Above all

Install NewBot CLI to take advantage of command lines

npm install -g newbot-cli

Create a project

newbot new <your directory name>

Go to the generated folder

You can test the chatbot with the command newbot serve. See below

Create main skill

main.converse is the conversational script of the main skill. It’s just the departure conversation

The language used is specific to NewBot. It is called ConverseScript and allows you to create conversational scripts with a syntax close to Javascript
More : https://newbot.io/docs/syntax/variables.html

@Event('start')
start() {
    > I am PizzaBot, and your name ?
}

Simple, isn't it? The chatbot gives its name from the first interaction with the user

start() is a function representing a dialog
@Event('start') is a decorator indicating how the function should be triggered. Here, the “start” event indicates that the function is triggered as soon as the first interaction with the chatbot occurs

Entering a text

Let's complete the main.converse file

$name = ''

@Event('start')
start() {
    > I am PizzaBot, and your name ?
    Prompt()
    $name = :text
    > Welcome, { $name }
}

$name is a global variable specific to the user
Prompt() is a native function waiting for the user to enter
:text a magic variable. It was created by the system. It contains the user entry. We store the value in the global variable $name

The Javascript file

The main.js file contains an object representing the skill

import code from './main.converse'

export default {
    code
}

Create a skill to place an order

First, create a new skill

newbot generate skill order

An order folder has been created in the bot/skills folder

order.converse

@Intent('order', [
    'order a pizza',
    'get pizza',
    'buy pizza'
])
order() {
    date = :intent.date.value
    callApi(date)
    > I have noted an order for { date }
}

We use an @Intent decorator to use the NLP

@Intent() is a decorator who defines an intention. First parameter: the name of the intention. Second parameter: the training phrase array
:intent.date.value is a magic variable that retrieves the date entered by the user
callApi() is a function created by us, present in Javascript

and JS skill

import code from './order.converse'

export default {
    code,
    functions: {
        callApi(date) {
            console.log(`call yout custom api, date is ${date}`)
        }
    }
}

The functions property allows you to declare functions that can be used in the conversational script

Link the created skill to the chatbot

Now that the skill order has been created, add it to the property skills in the main.js file

order.js

import code from './main.converse'
import orderSkill from './skills/order/order'

export default {
    code,
    skills: {
        orderSkill
    }
}

Full code

Test in Emulator

Type the following command at the root of your project

newbot serve

Go to localhost:3000 and test the project in the emulator

Go further in understanding

Deploy in browser : https://newbot.io/en/docs/essential/browser.html
Develop ExpressJS server and execute chatbot : https://newbot.io/en/docs/engine/express.html