Telegram Bot Demo with Scala/ Play

This page was originally created for a JUGs workshop.

The workshop is about the first steps on implementing a chat bot with:

• Telegram a Messaging app.
• telegrambot4s a Scala API for Telegram.

The following chapters document the workshop step by step.

A basic understanding of SBT and Scala will help;).

Setup the Bot

• Register a user on Telegram if you don't have one.
• Search for the BotFather in the Telegram App (he helps you to setup and manage your bots).
• Ask the BotFather for a new bot: /newbot
• He will guide you through the simple process (nice example of a chat bot;).
• The generated token is all what you need!

Setup the Project

• You need a SBT installation
• Clone this repo: git clone https://github.com/pme123/play-scala-telegrambot4s.git

The master branch contains only:

• Giter8 template for generating a Play project seed in Scala.
• the dependency: libraryDependencies += "info.mukel" %% "telegrambot4s" % "3.0.14"

JSON-Client

(Here the Solution Branch if you have problems)

Let's use the provided JSON-API from Telegram directly. We only wan't to show the info about our bot on the index page.

First we want to provide the token in a safe way to avoid leaking it.

• Create a package object in the bots package and add:

lazy val botToken: String = scala.util.Properties
  .envOrNone("BOT_TOKEN")
  .getOrElse(Source.fromResource("bot.token").getLines().mkString)

• You can now add a bot.token with your token and put it in your conf folder.
• (or if security is of no concern, you can add it directly: lazy val botToken = "[BOT_TOKEN]"

Implement the web-service client:

• Add a dependency for the web-service client: libraryDependencies += ws
• Inject the web-socket client and the execution context (async processing) in the HomeController:

class HomeController @Inject()(cc: ControllerComponents
                               , ws: WSClient)
                              (implicit ec: ExecutionContext)

• The URL for the API is needed: lazy val url = s"https://api.telegram.org/bot$botToken/getMe"
• Adjust the HomeController.index() function:

  def index() = Action.async { implicit request: Request[AnyContent] =>
    ws.url(url)
      .get() // as this returns a Future the function is now async
      .map(_.json) // the the body as json
      .map(_.toString())
      .map(str => Ok(views.html.index(str))) // forward the result to the index page
  }

• Adjust the index.scala.html with the added parameter and render it:

@(myBot: String)

@main("Telegram Bots rock!") {
  <h1>Welcome to our Bot!</h1>
    <p>@myBot</p>
}

• Open the sbt console in the project base: play-scala-telegrambot4s>sbt
• Run the application [play-scala-telegrambot4s] $ run
• In a browser go to: localhost:9000
• As result you should see a JSON string, like: {"ok":true,"result":{"id":301276637,"is_bot":true,"first_name":"MeetupDemoBot","username":"MeetupDemoBot"}}

That's it - let's move now to the Scala API. For the next Bot examples we don't need the Play server - so you can shut it down.

(you need to update the HomeControllerSpec - or if you are lazy remove it)

Webhooks vs Polling

Before we start let's explain shortly the difference between Webhooks and Polling. This is the quote from the Scala API:

Both methods are fully supported. Polling is the easiest method; it can be used locally without any additional requirements. It has been radically improved, doesn't flood the server (like other libraries do) and it's pretty fast.

Using webhooks requires a server (it won't work on your laptop). For a comprehensive reference check Marvin's Patent Pending Guide to All Things Webhook.

So for this workshop, or examples in general Polling is the way to go.

Hello User Bot

Let's greet the Bot and it should return that with a personalized greeting.

• Create a Scala object in the bots package: HelloBot

object HelloBot
  extends TelegramBot // the general bot behavior
    with Polling // we use Polling
    with Commands { // and we want to listen to Commands

  lazy val token: String = botToken // the token is required by the Bot behavior

  onCommand('hello) { implicit msg => // listen for the command hello and
    reply(s"Hello ${msg.from.map(_.firstName).getOrElse("")}!") // and reply with the personalized greeting
  }
}

• As we don't no infrastructure all we need is to run our Bot:

object BotApp extends App {
  HelloBot.run()
}

• Next we need our friend BotFather to create our command: /setcommands
• The easiest way is to manage your commands in file, from where you can take them (you always overwrite them for a bot). e.g. botcommands.txt:

hello - Simple Hello World.

• Like before the BotFather will help with this.
• Now lets say hello to your Bot (the command should be available, when hitting / in the text-field).

Counter Bot

(Here the Solution Branch if you have problems)

Callbacks

The first step to implement a conversation with a user is to understand the concept of callbacks. To guide the user through a conversation you can provide a keyboard. These keys (buttons) are identified with a callback identifier.

Create a Scala class CounterBot in the bots package (you can copy the HelloBot:

• We will listen for the Command /counter to start the process:

  onCommand("/counter") { implicit msg =>
    reply("Press to increment!", replyMarkup = Some(markupCounter(0)))
  }

• The logic for the counting and the creation of the button:

  private def markupCounter(n: Int): InlineKeyboardMarkup = {
    requestCount += 1
    InlineKeyboardMarkup.singleButton( // set a layout for the Button
      InlineKeyboardButton.callbackData( // create the button into the layout
        s"Press me!!!\n$n - $requestCount", // text to show on the button (count of the times hitting the button and total request count)
        tag(n.toString))) // create a callback identifier
  }

• The callback identifier is composed by a static tag and the actual count:

  private val TAG = "COUNTER_TAG"
  private def tag: String => String = prefixTag(TAG)

• When the user hits the button we can listen for it:

  onCallbackWithTag(TAG) { implicit cbq => // listens on all callbacks that START with TAG
    // Notification only shown to the user who pressed the button.
    ackCallback(Some(cbq.from.firstName + " pressed the button!"))
    // Or just ackCallback() - this is needed by Telegram!

    for {
      data <- cbq.data //the data is the callback identifier without the TAG (the count in our case)
      Extractors.Int(n) = data // extract the optional String to an Int
      msg <- cbq.message
    } /* do */ {
      request(
        EditMessageReplyMarkup( // to update the existing button - (not creating a new button)
          Some(ChatId(msg.source)), // msg.chat.id
          Some(msg.messageId),
          replyMarkup = Some(markupCounter(n + 1))))
    }
  }

• Like before we need a runner app:

object CounterBotApp extends App {
  CounterBot.run()
}

• we want to reuse our bot, so we only overwrite the command to counter - Counts the time a User hits the button.like above. As the commands are set always in one step it makes sense to manage them in file. Create 'bot-commands.txt` file and add:

  hello - Simple Hello World.
  counter - Counts the time a User hits the button.

• Run the CounterBotApp and select the command /counter
• Hit the button and create new buttons width /counter

Next steps (Next workshop;)

This was the basic workshop. Now we want to do complexer conversations. To get to this next level we need quite some ingredients:

• Handle the state of each user.
• Create a FSM (finite state machine) to design the conversation.
• A running App that easily integrates everything - and in a later state provides the webhooks.

Here starts the setup with Play to make more sense. As we use and integrate the FSM provided with Akka.