Getting started

A web application starts off with an HTTP server that creates a request and an empty response object and hands it over to the application framework, e.g. Express or Koa to generate a response.

A ChatAlly chat application is very similar: It starts off with a chat server that creates a request and response object and hands it over to the ChatAlly application. The application dispatches it to all the middleware that should generate the response and returns it to the server to send it out to the user client.

Installation

To get started, with the simplest possible chat application, you need create an npm ESM project, install the core library, a chat server, e.g. the ConsoleServer and implement some middleware.

npm init esnext -y
npm install @chatally/core @chatally/console

Create the application

Create a file index.js in which you configure your application

import { Application } from '@chatally/core'
import { ConsoleServer } from '@chatally/console'

new Application({ log: false }) //
  .use(new ConsoleServer())
  .use(({ req, res }) => {
    if (res.isWritable) {
      res.write(`You said '${req.text}'`)
    }
  })
  .listen()

Here we create a new chat application, register the ConsoleServer and implement a very simple middleware, that just echoes the user input.

ChatAlly is fully typed, so you can get full support from your code editor (you might need to setup a tsconfig.json file). See the guide Writing Middleware for details on how to write your own middleware.

You can now start the application by running

node index.js

Adding middleware

Of course, this chatbot is quite boring, but you can make it more interesting by adding some middleware, e.g. a module for NLP (Natural Language Processing) like nlp.js.

Install the module

npm install @chatally/nlpjs

and add it to your configuration instead of the function echo

import { Application } from '@chatally/core'
import { ConsoleServer } from '@chatally/console'
import { nlpjsMiddleware, trainNlp } from '@chatally/nlpjs'

const app = new Application({ log: false })
const nlp = await trainNlp(app.getLogger('nlp.js'))

app //
  .use(new ConsoleServer())
  .use(nlpjsMiddleware(nlp))
  .listen()

The function trainNlp creates a trained NLP module and expects a training corpus at corpus.json. For this example you can download a corpus.json from nlp.js’ Github. We integrate nlp.js’ logging with ChatAlly’s by passing a logger from the application to nlp.js.

The function nlpjsMiddleware wraps the nlp.js Nlp.process(...) method in a ChatAlly middleware function.

Now it should be a bit more fun talking to your chatbot.

What next 👉

You could now improve your application and change the server to one from the servers section in the sidebar. You could add more or different middleware, explore the middleware section in the sidebar. Deploy your application, or start writing your own middleware to add any missing piece.

If you first want to learn more about the core concepts, check out the reference documentation.