Important
This tutorial is outdated, and no longer works because the Python Telegram Bot library was changed.
In this tutorial you're going to learn how to use Python to make a telegram bot.
For this tutorial you need a computer with Python 3.7 or higher installed. You can download it directly from the Python website.
Pay attention that on Windows the Python installer must be run as Administrator to correctly install pip
into your PATH
to simplify the next step.
Next you'll need to install the python-telegram-bot
library that provides most of the Telegram bot functionality. If your environment is set up correctly you can do that by running the command:
pip install python-telegram-bot
Finally you'll need a programming environment. While you can use any text editor and terminal combination, In this tutorial I use Visual Studio Code. Once you have VSCode installed, click the Extensions menu or press CtrlShiftX and search for and install the Python extension by Microsoft.
Next you're going to have to download this project's start point. Extract the ZIP somewhere on you computer.
Now open VSCode and select File ยป Open Folder and open the folder that you just extracted.
Open the file bot.py
and press F5. If everything went well, you should see a message that says:
Exception has occurred: InvalidToken
That means the program starts correctly and you're ready to create your Telegram bot.
To create your own bot you'll need to do two things. You need to register a bot with Telegram. Then you need to make the Python program connect to Telegram and log in as that bot.
To create a telegram bot, search for Botfather on Telegram and enter a conversation with him.
Botfather is the original Telegram bot and talking to him is currently the only way to create new Telegram bots.
Use the /newbot
command to start the process of creating a new bot.
Next enter a display name for your bot. I'm going to use BUK Telegram tutorial.
Then enter a username for the bot. Keep in mind that it must end with bot
.
Once you've done that, the Botfather will give you a Token.
Keep the token secret! This token acts as a combined username and password for your bot. If someone else has it, they can impersonate your bot. If someone has stolen your token, you can talk to the Botfather and generate a new token. This is like a password change. The old token stops working, and the Bothfather gives you a new token for your bot.
This is why I can safely show my token in the video tutorial. I have already replaced it with a new one after recording the video.
You might have noticed that the Python program contains a line that says:
TOKEN = 'YOUR TOKEN HERE'
Insert your token there between the quotes.
Now that you have registered you bot and added the token to the program, test that it works.
Press F5 to start the bot program.
Search for your bot in Telegram. Once you find it, open the conversation and click the Start button. This is actually a security feature. No bot is allowed to talk to you unless you Start to talk to it first.
Once you have started the conversation, you can now send /hello
to the bot. If everything works correctly, the bot should reply.
Talk to the Botfather and tell him /setcommands
. He'll ask you to specify the commands in a specific format. In this case send:
hello - A simple greeting
Now when you click the / button in Telegram the /hello
command will be available with explanation.
It is not necessary, nor always practical, to register commands, but it can be nice to users of your bot.
To make a bot useful, it needs to actually provide some interesting service. One of the reasons this tutorial is a bit more advanced, is that you'll have to decide for yourself what exactly you bot is going to do. However before you can do that, you'll need to understand the basics of how to handle Telegram messages.
Having a look at the current program, you'll see the line:
updater.dispatcher.add_handler(CommandHandler('hello', hello))
Here we add a handler. In this case we add a command handler that runs the hello
function when the user sends the command hello
to the bot.
If you want to know more about what handlers are available, check documentation of the telegram-bot-api library.
If you're serious about writing a Telegram bot, you're going to spend some time digging through that documentation. However, that is very technical documentation. Once you really are going to write your own thing I can recommend first looking at:
- The python-telegram-bot examples
- The python-telegram-bot code snippets on the wiki
For now let's instead focus on what we can do inside the handler's function. The current hello function looks like this:
def hello(update: Update, context: CallbackContext) -> None:
update.message.reply_text(f'Hello {update.effective_user.first_name}')
What you can see here is that you receive two things in the handler. The update
and the context
. In this case we only use the update
to send a textual reply using the update.message.reply_text
function. The text we send is Hello
followed by the value of update.effective_user.first_name
. Which is the first name of the user that sent the hello command that caused this function to start.
Let's add a different handler. The basic MessageHandler handles any kind of message, but we can filter those to only handle specific messages. In this case, we're going to add a handler that simply echoes back whatever the user said.
For this we need to do three things:
- Define an echo function
- Handle messages using the echo function
- Import the necessary parts from the library to make it work
First we'll write the echo function, you can place it below the hello
function:
def echo(update: Update, context: CallbackContext) -> None:
update.message.reply_text(update.message.text)
As you can see, it simply replies with the text value of the received message.
Next we'll have to make sure that we run this function when we receive a message that contains text and is not a command. Add this below the line where we add the hello CommandHandler:
updater.dispatcher.add_handler(MessageHandler(Filters.text & ~Filters.command, echo))
This adds a message handler that executes the echo
function but only for incoming messagest that match the filters text
and not command
.
Finally to make this actually work, we need to immport the correct parts of the library.
Change the import from telegram.ext
line near the top of the file, to read:
from telegram.ext import Updater, CommandHandler, CallbackContext, MessageHandler, Filters
Now start you bot and see whether it works. If you did it correctly, whenever you send it /hello
it will reply as it did before. When you send it something else, it will repeat that text back.
An echo bot is nice, but we can do better. We can make the bot sarcastic.
Let's change the echo function to:
def echo(update: Update, context: CallbackContext) -> None:
update.message.reply_text(
''.join([l.upper() if i % 2 == 0 else l for (i, l) in enumerate(update.message.text)])
)
Now that's quite a compact piece of code, that warrants some explanation.
enumerate(update.message.text)
changes the input text into a list of pairs. These pairs consist of an index and the letter. So the text echo
is turned into [(0, 'e'), (1, 'c'), (2, 'h'), (3, 'o')]
.
The for (i, l) in enumerate...
part states that we're going to loop over those pairs. In each pair we call the first entry, the number, i
and the second entry, the letter, l
.
l.upper() if i % 2 == 0 else l
is run for each of those i, l
pairs and checks whether i
is even by using % 2
to get the remainder after division by two, then using == 0
to ensure the remainder is zero. If that is true, it turns l
into an upper case letter. else l
means that if i
is odd, l
is not changed.
This means that [(0, 'e'), (1, 'c'), (2, 'h'), (3, 'o')]
is turned into ['E', 'c', 'H', 'o']
. Finally the surrounding ''.join(...)
takes the list of letters we now have and turns them back into a single line of text by putting ''
(nothing) in between each part of the list. Hence ['E', 'c', 'H', 'o']
is turned into EcHo
.
Now you have made you first bot that actually does something with the information the user sent you. Give it a try!
One nice feature of Telegram is that you can send people buttons that they can click. Another nice feature is that you can send an emoji to make a random number. Let's combine those to make a command that lets you choose the random number emoji to use.
First the random
function. This function sends you a text reply that asks you to choose one of the options.
It also shows an InlineKeyboardMarkup
which contains several InlineKeyboardButton
s. On the first row we have a button that reads Basketball and whose value is the basketball emoji.
On the second line we have first a button for Dice whose value is a die emoji and next a button for Darts whose value is a bullseye emoji.
def random(update: Update, context: CallbackContext) -> None:
reply_buttons = InlineKeyboardMarkup([
[InlineKeyboardButton("Basketball", callback_data='๐')],
[
InlineKeyboardButton("Dice", callback_data='๐ฒ'),
InlineKeyboardButton("Darts", callback_data='๐ฏ'),
]
])
update.message.reply_text(
f'Hello {update.effective_user.first_name}, please choose an option:',
reply_markup=reply_buttons
)
Don't forget to add the command handler:
updater.dispatcher.add_handler(CommandHandler('random', random))
Your imports should start with:
from telegram import Update, InlineKeyboardButton, InlineKeyboardMarkup
If we try this now, we can see the buttons, clicking them shows a clock icon on the clicked button, but nothing happens. We should add a handler for the button click.
Once such a button is pressed, the result should be handled in another function. Here we first mark the question we posed by showing the buttons as answered by calling update.callback_query.answer()
. Telegram requires that we do this.
Next we edit the original message where we showed the buttons, and we replace the list of buttons with an empty list. This removes the buttons, so that the user can only click them once.
Finally we reply to the button click by sending the emoji value of the button that was pressed using send_dice
to use it as a random number generation animation.
def button(update: Update, context: CallbackContext) -> None:
# Must call answer!
update.callback_query.answer()
# Remove buttons
update.callback_query.message.edit_reply_markup(
reply_markup=InlineKeyboardMarkup([])
)
update.callback_query.message.reply_dice(emoji=update.callback_query.data)
Don't forget to add this handler:
updater.dispatcher.add_handler(CallbackQueryHandler(button))
And make sure you have all these imports:
from telegram import Update, InlineKeyboardButton, InlineKeyboardMarkup
from telegram.ext import Updater, CommandHandler, CallbackContext, MessageHandler, Filters, CallbackQueryHandler
Give it a try. First send /random
. Then click one of the buttons. If everything went well, the buttons are replaced by an animated random number of your choice.
Now we have learned to talk back and forth about with the bot. However it would be nice if the bot could reply based on something we told it before.
The library we're using provides the context.user_data
for this purpose. Let's make use of it.
This function checks what it knows about the user, and lists it. Then it asks for more information.
def personal(update: Update, context: CallbackContext) -> int:
reply_list = [f'Hello {update.effective_user.first_name}']
if context.user_data:
reply_list.append('I know these things about you')
reply_list.extend([f'Your {key} {value_pair[0]} {value_pair[1]}' for (key, value_pair) in context.user_data.items()])
else:
reply_list.append('I don\'t know anything about you.')
reply_list.extend([
'Please tell me about yourself.',
'Use the format: My X is/have/are Y'
])
update.message.reply_text('\n'.join(reply_list))
Let's run that command whenever the user starts talking to the bot or runs the /start
command:
updater.dispatcher.add_handler(CommandHandler('start', personal))
If you try it out, you'll quickly realise that you only get sarcastic remarks back from your bot. We haven't actually written functionality that lets us tell the bot something about ourselves.
This function works on messages matching the given Regular expression. Which is a method of expressing very simple grammars for text. The INFO_REGEX
matches sentences such as My head is large, My parents have children, My shoes are too small. Besides that, thanks to the placement of the brackets, it will save the parts inside the brackets as capture groups. Hence the sentence My shoes are too small yields these groups: `['shoes', 'are', 'too small']
We save this to the context.user_data
in such a way that the sentence My hands are too large results in the data being saved as if we executed: context.user_data['hands'] = ('are', 'too large')
INFO_REGEX = r'^My (.+) (is|have|are) (.+)$'
def receive_info(update: Update, context: CallbackContext) -> int:
# Extract the three capture groups
info = re.match(INFO_REGEX, update.message.text).groups()
# Using the first capture group as key, the second and third capture group are saved as a pair to the context.user_data
context.user_data[info[0]] = (info[1], info[2])
# Quote the information in the reply
update.message.reply_text(
f'So your {info[0]} {info[1]} {info[2]}, how interesting'
)
Now we add a handler that runs this function, but the filter specifies that the function should only be executed for messages that match the INFO_REGEX
.
Important Put this line before the line that adds the sarcastic reply handler. Because the first matching handler is used. And the sarcastic reply matches any text that is not a command.
updater.dispatcher.add_handler(MessageHandler(Filters.regex(INFO_REGEX), receive_info))
Finally to make regular expressions work we must import that functionality by adding this line to imports:
import re
Start you bot, and if everything went well, you can run /start
and it will tell you that it knows nothing about you. You can then tell it things about you.
If you run /start
again, he will reply with what he knows about you.
One last thing that you might have noticed is that the bot does not remember this information across restarts. Luckily the library we're using has us covered. Persistence is a built-in feature and extremely easy to add.
Find the line that says:
updater = Updater(TOKEN)
Change it to:
updater = Updater(TOKEN, persistence=PicklePersistence(filename='bot_data'))
Of course PicklePersistence
must be imported. Your import line should look like this:
from telegram.ext import Updater, CommandHandler, CallbackContext, MessageHandler, Filters, CallbackQueryHandler, PicklePersistence
Start the bot again, and tell it some things about yourself. Stop it, and restart it. If you run the /start
command, it should immediately list the things you told it before the restart.
To learn about conversations, I recommend the persistent conversation example example.
It combines most of what you have learned in this tutorial with the ConversationHandler
which is a handler that combines other handlers with basic state management to create a conversation with multiple stages.
There are plenty of other interesting examples to try there too.
If you want to upload your code somewhere, such as here to Github, you don't want to upload your token. Python provides a fairly simple mechanism to do this. Create a new file that is named .env
next to your bot's python file. Find the line of code that says:
TOKEN='YOUR TOKEN HERE'
Move that line from the Python program to the .env
file. If you have spaces around the equals sign (e.g. =
) please remove them. Then find the line in the program that says:
updater = Updater(TOKEN, persistence=PicklePersistence(filename='bot_data'))
Replace TOKEN
with os.environ['TOKEN']
. os.environ
contains environment variables. When you run a Python program and there is an adjacent .env
file, it will add all X=Y
key-value pairs from the .env
file. Therefore os.environ
will have your token.
updater = Updater(os.environ['TOKEN'], persistence=PicklePersistence(filename='bot_data'))
Finally ensure the program imports os
. Add this to your import lines:
import os
Finally I'm using git
for version control. Git has an ignore mechanism allows us to exclude files from version control. Therefore create a new file called .gitignore
containing:
bot_data
.env
This ensures that the bot_data
and the .env
file containing the token are never saved in version control.
Run you bot again and verify that it still works. It now loads the token from the .env
file.
Now you've noticed that your bot only works as long as the Python program is running on your computer. Once you've got it working and want it to be available to anyone, you might want to deploy your bot to a cloud service that runs it for you.
Warning Heroku no longer offers the service described here for free! I recommend you look for another way to deploy your bot.
In this case I'm going to put the code on Github and deploy to Heroku. You'll need two things for this:
Once you have those set up, you can got to the Source Control (CtrlShiftG) tab in VSCode and press the Publish to GitHub button. You'll be prompted to login to GitHub.
It will ask you whether you want to publish this as a private or as a public repository. Now if you put your token in the .env
file and added the .gitignore
file, you can safely choose public. A public repository can be seen by everyone on the internet. A private repository can be seen only by you.
If you accidentally published your token. You can talk to the Botfather and ask him to /revoke
the token. You'll get a new token for your bot.
Now login to Heroku. Click New ยป Create new app. Choose a name for your app, and choose whether to host in the US or in Europe. For Telegram bots such as these, it isn't that important, in general it's easiest to choose the closest location to your customers.
On the next page in the Deployment method section, choose GitHub and complete the steps to connect your GitHub account. Once you've done that, hit the Search button to list all your GitHub repositories and choose the telegram bot one.
Once you've done that, a little bit lower is the option for Automatic deploys. Press Enable Automatic Deploys. Now once you change your bot and push that update to GitHub, Heroku will automatically update your bot.
Finally we must tell Heroku the token, which is not in the code on GitHub, because we don't want to publish that. Go to the Settings tab in Heroku and look for the Config Vars section. Click Reveal Config Vars. In the Key field enter TOKEN
and in the Value field enter your token. Then press Add. Now Heroku knows your token and it will be available in the environment where your bot runs. Therefore the os.environ['TOKEN']
in Python will be your token, even if we put it in the environment using the Heroku configuration, instead of the .env
file.
Now one last thing we should do to make the Heroku deployment work, is push a change to the repository. Luckily we must make a few changes to make the app work on Heroku.
You might remember that at the very start you had to pip install python-telegram-bot
to download the library. We must tell Heroku that it should do that. Python has a standard mechanism for that, and if we use that, as an added bonus Heroku will automatically detect that our program is written in Python.
Add a new file called requirements.txt
it should contain this line:
python-telegram-bot
Next we must tell Heroku how it should start the program. Heroku's standard for this is the Procfile
. Create a new file called Procfile
(mind the capital P, procfile
won't work). It should contain:
worker: python3 bot.py
Go to the source control tab of VSCode again. You'll see that the added fiiles are shown. Click the +
icon after every file name. Then enter a commit message for example: Add Heroku configuration. Confirm it by clicking the Commit check-mark that appears when you hover over the text SOURCE CONTROL or by pressing CtrlEnter.
Finally in the bottom left corner of your VSCode window you should see the GitHub branch and status. Click on the synchronisation icon ๐ to push your changes to GitHub.
Now your code will be auto-deployed to Heroku.
Go to you app's Overview on Heroku. You'll see a Dyno formation menu. Click Configure dynos. If deployment went well, you'll see the worker python3 bot.py
deployment here. Click the edit pencil behind that worker. Make sure your bot is not running on your local computer! Now click the switch on Heroku to activate the dyno and click Confirm.
Start talking to your bot on Telegram. If everything went well, you'll notice that it responds because it is running in the cloud.