Building a Chatbot

Building a Chatbotusing Telegram and PythonDiogo RibeiroBlockedUnblockFollowFollowingMay 28Photo by Lance Anderson on UnsplashChatbots are all the rage at the moment, with some predicting that they will be bigger than mobile apps.

The main idea of chatbots is that instead of having to dig through awkward mobile menus and learn UIs, you’ll simply have a conversation with a bot through a familiarinstant messaging interface.

If you want to order a Pizza, you start a conversation with the Domino’s Pizza bot and have the same conversation with it that you might have with a human.

There are a few different platforms that allow you to build your own chatbot.

One of these, which is arguably the simplest to use and is also growing steadily in popularity, is Telegram.

In this tutorial, we’ll walk through building a simple Telegram Bot using Python.

At first, our bot will simply echo back any message we send it, but then we’ll extend it to add a database and persist information across chat sessions.

We’ll use Python to power our Bot and SQLite to store information persistently across sessions.

In summary, this is a tutorial series that will:Show you how to write a simple Echo Bot from scratch using Python and the Telegram Bot API (Part 1)Extend the Echo Bot into a ToDo list manager bot, backed by a SQLite database (Part 2)Show how to run our Bot from a VPS and allow it to scale to more users (Part 3).

Although creating an Echo Bot is simple enough, and you can find various scripts and frameworks online that will give you this as a starting point — we will do everything from scratch and explain every piece of code we write.

We’ll also look at some subtleties in the Telegram API, and talk about what these mean for us as developers.

If you just want to create a Telegram bot as quickly as possible, this tutorial is probably not what you’re looking for, but if you want to gain a deeper understanding of how chatbots work and how to build one from scratch, then you’re in the right place.

What you needYou’ll need to have:Some basic Python knowledge to follow this tutorialYou should be comfortable with running commands in a Linux Shell, a MacOS Terminal, or a Windows Command PromptYou should be able to install Python packages using the pip package manager (or conda if you’re more comfortable with that)Ideally, you should have written at least a basic SQL statement before, but this is not strictly necessary (and will only be relevant in Part 2).

All of the code is aimed at Python 3.

5, but it should be easily adaptable to other versions of Python.

Why Python?You can write a Telegram chat bot in any language you want.

Some of the main options apart from Python would be Java, PHP, or Ruby.

If you are more familiar with a different high-level programming language, then you might prefer to use that instead, but Python is a good choice for several reasons:Python can make HTTP requests very concisely and simply through the requests module.

Getting the content from a URL (which is how we’ll be controlling our Telegram Bot) would need many more lines of Java than the Python equivalent.

Python is the most popular language for natural language processing and machine learning: although we won’t be using either of these for our simple bot, both of them would be necessary for a more advanced Bot.

Thus, if you want to extend the Bot, it’s good to get comfortable with Python.

Python has good support for serving web content: when we want to scale up our Bot to allow it to receive many messages per second, Python has mature technologies such as WSGI to reach “web scale”.

Python is portable — we can easily run the same code on Linux, MacOS, or Windows.

Setting upNearly everything we do will be achievable using only the standard Python libraries, but we’ll also be using the third-party requests module which provides a number of improvements to Python's urllib, and allows us to make HTTP requests very simply and concisely.

Install this through pip using a command similar to the following (you may need to use pip instead of pip3 and/or add the –user flag, based on how you usually install Python libraries).

pip3 install requestsIf you usually use a virtual environment for new Python projects, then set one of those up first, and install requests inside that.

Creating a Telegram BotThe first step is to tell Telegram that you want to create a new bot.

All the messages that our Bot sends and receives will go through Telegram’s infrastructure.

Our code will periodically make a request to retrieve all new messages to our Bot from Telegram’s servers, and will then send responses to each message as necessary.

In order to register a bot with Telegram, you first need to create a personal Telegram account.

Visit web.


org and enter your phone number.

Telegram will send you a text message (SMS), and you can then create an account by following the instructions on the screen.

If you already have a Telegram account, then you can simply use that one, and you can also use any of the Telegram Desktop and Mobile apps available from telegram.

org, instead of the Web app that we’ll be using for all examples in this tutorial.

Once you have a Telegram account, you can register a new Telegram Bot by using Bot Father.

Visit telegram.

me/botfather to start a conversation with Telegram’s bot that creates other bots.

Telegram bots can receive messages or commands.

The former are simply text that you send as if you were sending a message to another person, while the latter are prefixed with a / character.

To create a new bot, send the following command to Bot Father as a chat (exactly as if you were talking to another person on Telegram).

/newbotYou should get a reply instantly that asks you to choose a name for your Bot.

We’ll call our Bot To Do Bot because, by the end of this tutorial, it'll function as a simple "to do" list.

Send the following message to Bot Father when it prompts you for a name:To Do BotBot Father will now ask you to pick a username for your Bot.

This username has to end in bot, and be globally unique.

As Telegram has grown more popular, it has become more difficult to find a short and relevant username for your Bot.

In this tutorial, we'll be using exampletodo_bot, but for the rest of this tutorial, we'll indicate the Bot's username with <your-bot-username>, so you'll have to substitute your chosen username wherever relevant from now on.

Send your chosen username to Bot Father:<your-bot-username>Now Bot Father will send you a “Congratulations” message, which will include a token.

The token should look something like this:2483457814:AAHrlCx234_VskzWEJdWjTsdfuwejHyu5mIFor the rest of this tutorial, we’ll indicate where you need to put your token by using <your-bot-token>.

Take note of the token, as we’ll need it in the code that we’re about to write.

Interacting with our Bot through our web browserWe can control our Bot by sending HTTPS requests to Telegram.

This means that the simplest way to interact with our Bot is through a web browser.

By visiting different URLs, we send different commands to our Bot.

The simplest command is one where we get information about our Bot.

Visit the following URL in your browser (substituting the bot token that you got before)https://api.


org/bot<your-bot-token>/getmeThe first part of the URL indicates that we want to communicate with the Telegram API (api.



We follow this with /bot to say that we want to send a command to our Bot, and immediately after we add our token to identify which bot we want to send the command to and to prove that we own it.

Finally, we specify the command that we want to send (/getme) which in this case just returns basic information about our Bot using JSON.

The response should look similar to the following:{"ok":true,"result":{"id":248718785,"first_name":"To Do Bot","username":"exampletodo_bot"}}Retrieving messages sent to our BotThe simplest way for us to retrieve messages sent to our Bot is through the getUpdates call.

If you visit https://api.


org/bot<your-bot-token>/getUpdates, you'll get a JSON response of all the new messages sent to your Bot.

Our Bot is brand new and probably hasn't received any messages yet, so if you visit this now, you should see an empty response.

Telegram Bots can’t talk to users until the user first initiates a conversation (this is to reduce spam).

In order to try out the getUpdates call, we'll first send a message to our Bot from our own Telegram account.

Visit telegram.

me/<your-bot-username> to open a conversation with your Bot in the web client (or search for @<your-bot-username> in any of the Telegram clients).

You should see your Bot displayed with a /start button at the bottom of the screen.

Click this button to start chatting with your Bot.

Send your Bot a short message, such as "hello".

Now visit the https://api.


org/bot<your-bot-token>/getUpdates URL again, and you should see a JSON response showing the messages that your bot has received (including one from when you pressed the start button).

Let's take a look at an example of this and highlight the import data that we'll be writing code to extract in the next section.

{"ok":true,"result":[{"update_id":625407400,"message":{"message_id":1,"from":{"id":24860000,"first_name":"Gareth","last_name":"Dwyer (sixhobbits)","username":"sixhobbits"},"chat":{"id":24860000,"first_name":"Gareth","last_name":"Dwyer (sixhobbits)","username":"sixhobbits","type":"private"},"date":1478087433,"text":"/start","entities":[{"type":"bot_command","offset":0,"length":6}]}},{"update_id":625407401,"message":{"message_id":2,"from":{"id":24860000,"first_name":"Gareth","last_name":"Dwyer (sixhobbits)","username":"sixhobbits"},"chat":{"id":24860000,"first_name":"Gareth","last_name":"Dwyer (sixhobbits)","username":"sixhobbits","type":"private"},"date":1478087624,"text":"test"}}]}The result section of the JSON is a list of updates that we haven't acknowledged yet (we'll talk about how to acknowledge updates later).

In this example, our Bot has two new messages.

Each message contains a bunch of data about who sent it, what chat it is part of, and the contents of the message.

The two pieces of information that we'll focus on for now are the chat ID, which will allow us to send a reply message and the message text which contains the text of the message.

In the next section, we'll see how to extract these two pieces of data using Python.

Sending a message from our BotThe final API call that we’ll try out in our browser is that used to send a message.

To do this, we need the chat ID for the chat where we want to send the message.

There are a bunch of different IDs in the JSON response from the getUpdates call, so make sure you get the right one.

It's the id field which is inside the chat field (24860000 in the example above, but yours will be different).

Once you have this ID, visit the following URL in your browser, substituting <chat-id> for your chat ID.



org/bot<your-bot-token>/sendMessage?chat_id=<chat-id>&text=TestReplyOnce you’ve visited this URL, you should see a message from your Bot sent to your which says “TestReply”.

Now that we know how to send and receive messages using the Telegram API, we can get going with automating this process by writing some logic in Python.

Writing the Python code for our BotNow we can get to writing Python.

Create the file echobot.

py and add the following code:import json import requestsTOKEN = "<your-bot-token>"URL = "https://api.



format(TOKEN)def get_url(url): response = requests.

get(url) content = response.


decode("utf8") return contentdef get_json_from_url(url): content = get_url(url) js = json.

loads(content) return jsdef get_updates(): url = URL + "getUpdates" js = get_json_from_url(url) return jsdef get_last_chat_id_and_text(updates): num_updates = len(updates["result"]) last_update = num_updates – 1 text = updates["result"][last_update]["message"]["text"] chat_id = updates["result"][last_update]["message"]["chat"]["id"] return (text, chat_id)def send_message(text, chat_id): url = URL + "sendMessage?text={}&chat_id={}".

format(text, chat_id) get_url(url) text, chat = get_last_chat_id_and_text(get_updates())send_message(text, chat)Let’s pull apart what this code does:In lines 1 and 2, we import the requests and json modules.

The first is to make web requests using Python and we'll use it to interact with the Telegram API (similarly to what we were using our web browser for earlier).

We'll use the JSON module to parse the JSON responses from Telegram into Python dictionaries so that we can extract the pieces of data that we need.

The next two lines are global variables, where we define our Bot’s token that we need to authenticate with the Telegram API, and we create the basic URL that we’ll be using in all our requests to the API.

The get_url function simply downloads the content from a URL and gives us a string.

We add the .

decode("utf8") part for extra compatibility as this is necessary for some Python versions on some platforms.

Normally, we'd do some exception handling here as this request could fail if our internet connection were down, if Telegram's service were down, or if there were an issue with our Token.

However for simplicity, here we'll simply assume that everything Always Works (TM).

The get_json_from_url function gets the string response as above and parses this into a Python dictionary using json.

loads() (loads is short for Load String).

We'll always use this one as Telegram will always give us a JSON response.

get_updates calls the same API command that we used in our browser earlier, and retrieves a list of "updates" (messages sent to our Bot).

get_last_chat_id_and_text provides a simple but inelegant way to get the chat ID and the message text of the most recent message sent to our Bot.

Because get_updates will always send all the messages that were recently sent to our bot, this is not ideal, as we will always download a whole bunch of messages when we only want the last one.

We'll discuss later in more detail how to do this more elegantly.

For now, this function returns a tuple of the chat_id which identifies the specific chat between our Bot and the person who sent the message, and the text, which is the message itself.

send_message takes the text of the message we want to send (text) and the chat ID of the chat where we want to send the message (chat_id).

It then calls the sendMessage API command, passing both the text and the chat ID as URL parameters, thus asking Telegram to send the message to that chat.

The final two lines bring everything we have written together to actually receive and send a message.

First, we get the text and the chat ID from the most recent message sent to our Bot.

Then, we call send_message using the same text that we just received, effectively "echoing" the last message back to the user.

At the moment, our script doesn’t listen for new messages and immediately reply.

Instead, when we run it, our bot will fetch only the most recent message sent to that and echo it.

We can test it out by sending our bot a message, and then running the script.

Give this a go!Flaws with our botThe most obvious problem with our Bot is that we have to run a Python script manually every time we want to interact with it.

Also, as mentioned before, we always download the entire message history that Telegram provides.

This is both inefficient and unreliable, as we don’t want to unnecessarily download the entire message history if we only want a single message, and because Telegram only keeps this list of updates for 24 hours.

Another issue is that we pass our message as a string, but because this is converted to a URL before being sent to Telegram, you’ll notice that some unexpected things happen if you send messages to the bot with special characters (for example, the + symbol will disappear from all echoed messages).

Finally, the Bot throws an index error if we try to run it when there are no new messages to receive.

We’ll now update our bot to:Constantly listen for new messages and reply to each.

Acknowledge each message as it receives it and tells Telegram to not send us that message again.

Use Long Polling so that we don’t have to make too many requests.

Correctly encode our messages to account for URL formatting.

Listening for new messagesWe don’t want to manually start our Bot every time that we want it to reply to the latest message, so the first thing to do is to wrap our code that receives new messages and echoes them back in a loop.

We’ll also put this in a main function and use the Pythonic if __name__ == '__main__' statement so that we could import our functions into another script without running anything.

We don't want to ask for new updates as fast as possible, so we'll also put a small delay between requests (this is kinder to Telegram's servers and better for our own network resources, too).

At the top of the file, add a new import for the Python time moduleimport timeAnd change the last two lines of the file to read as follows:def main(): last_textchat = (None, None) while True: text, chat = get_last_chat_id_and_text(get_updates()) if (text, chat) != last_textchat: send_message(text, chat) last_textchat = (text, chat) time.


5)if __name__ == '__main__': main()This code now gets the most recent messages from Telegram every half second.

We now also need to remember the most recent message that we replied to (we save this in the last_textchat variable) so that we don't keep on sending the echoes every second to messages that we've already processed.

This is again a very crude way of achieving what we want (for example, if we send the same message to our bot twice in a row, it won't reply to the second one), but we'll see a more elegant way to achieve this below.

For now, you can run this code and now instead of the script terminating, you'll see that it keeps running.

You can now send your Bot a series of messages, and (as long as you don't send more than one per half second), you'll see each of them getting echoed back again.

Acknowledging the messages we’ve already seenInstead of asking Telegram for all our recent messages with every call, and then trying to figure out which ones we are interested in, we can tell Telegram that we’ve already processed certain messages and that we want to stop receiving them as part of the getUpdates calls.

Each update has an update_id field, and these are incremental (later messages have higher numbers).

When we make the getUpdatesAPI call, we can optionally pass an offset argument and give an update_id as the value.

This tells Telegram that we've already seen and processed that message and that we don't want it again.

This also means that Telegram will never send us any of the previous messages (messages with a lower update_id) again either, so we need to make sure that we really are finished with all of the messages before doing this.

Modify our bot as follows:Add an optional offset parameter to our getUpdates function.

If this is specified, we'll pass it along to the Telegram API to indicate that we don't want to receive any messages with smaller IDs than this.

The modified function should look like this:def get_updates(offset=None): url = URL + "getUpdates" if offset: url += "?offset={}".

format(offset) js = get_json_from_url(url) return jsAdd a function that calculates the highest ID of all the updates we receive from getUpdates.

This should look as follows.

def get_last_update_id(updates): update_ids = [] for update in updates["result"]: update_ids.

append(int(update["update_id"])) return max(update_ids)This simply loops through each of the updates that we get from Telegram and then returns the biggest ID.

We need this so that we can call getUpdates again, passing this ID, and indicate which messages we've already seen.

Add a function to send an echo reply for each message that we receive.

This should look as follows:def echo_all(updates): for update in updates["result"]: try: text = update["message"]["text"] chat = update["message"]["chat"]["id"] send_message(text, chat) except Exception as e: print(e)Update the code in main() so that it looks like this:def main(): last_update_id = None while True: updates = get_updates(last_update_id) if len(updates["result"]) > 0: last_update_id = get_last_update_id(updates) + 1 echo_all(updates) time.


5)Our main code no longer needs to worry about duplicate messages, as each time we get new messages, we send the biggest update_id along with the next request, ensuring that we only ever receive messages that we haven’t seen before.

Note that we have to check if there are new updates (which we do in the third line of main()), and that we have to always send an update ID which is one bigger than the previous one we've seen (i.


we're actually telling Telegram which ID we're expecting, not which one we've seen).

Try out the changes by restarting the Python script and sending some messages to your Bot — you should see that it works as before, but now it doesn’t matter if you send duplicate messages or send messages too quickly, both of which are big improvements.

Using Long PollingThe last major problem with our Echo Bot is that it has to make a web request every 0.

5 seconds.

This is not great for Telegram’s servers (they explicitly ask people not to do this outside of testing scenarios) and not great for our resources either.

Long Polling takes advantage of the fact that most of the time, we are receiving “empty” responses.

Because our Bot is probably not going to be receiving messages every half second, most of the time when we ask for updates, there aren’t any.

With Long Polling, instead of Telegram telling us that there aren’t updates, it simply keeps the connection open until there are updates, and then sends these down the open pipe.

Of course, it’s impractical to keep a connection open forever, so we can specify the number of seconds that we want to wait for.

This is done by passing another optional argument to the getUpdates call, namely timeout.

To make our code use Long Polling, simply update our get_updates method as follows:def get_updates(offset=None): url = URL + "getUpdates?timeout=100" if offset: url += "&offset={}".

format(offset) js = get_json_from_url(url) return jsNow we always pass along the timeout argument.

Because we now have two arguments, we also need to change where we previously had ?offset={} to &offset={} (in URLs, we specify that the argument list is starting with a ? but further arguments are separated with &).

Run the bot again, and it should run exactly as before, but now it’ll be making far fewer requests and using less of your machine’s resources.

If you want to check that this is working, simply add a line like print("getting updates") directly below the while True in the main function and run the bot with and without the timeoutargument that we just added.

Without the timeout, you'll see that the bot checks for updates every 0.

5 seconds.

While with the timeout, it will only initiate a new check every 100 seconds, or whenever a new message is received.

Correctly encoding our message textThe final problem of our echo bot is that it acts strangely if we send it messages containing special characters.

For example, all + signs disappear from our messages, and all text after an & sign disappears, too.

This is caused by these symbols having special meanings in the context of URLs.

To fix this, we need to encode any special characters in our message.

Luckily, the standard Python urllib has a function that handles this for us, so we only need to import that and add a single line of code.

Add the following line at the top of your .

py fileimport urllibAnd now modify the send_message function to read as follows:def send_message(text, chat_id): text = urllib.


quote_plus(text) url = URL + "sendMessage?text={}&chat_id={}".

format(text, chat_id) get_url(url)Restart the Bot once more, and send it some messages that were problematic before, such as:+Hello+Hello&testNow it should be able to reply to all of these messages (and pretty much anything else you throw at it, including emoji) flawlessly.

Adding a SQLite Database BackendIn this tutorial, we’ll add a SQLite database backend to our bot and allow it to remember information for specific users indefinitely.

We’ll build a simple To Do list that allows users to add new items or delete existing ones.

In addition to the basic Python knowledge required for Part 1, it would be good if you had some familiarity with relational databases for this tutorial.

Ideally, you should already be familiar with concepts such as Tables, Rows, Columns, and have written at least a basic SELECT * FROM SQL statement.

But if you haven't, you'll probably be able to keep up anyway.

OverviewIn this tutorial we’ll cover the following:Building a Database Helper script to cover adding data to a SQLite database and retrieving it again.

Adding a custom Telegram Keyboard, so that our user can select options instead of typing out an entire message.

Adding the concept of ownership to items, so that multiple users can use our bot.

By the end of this tutorial, you’ll know how to use some of the more advanced features of the Telegram API (specifically, the custom keyboard feature), and how to add a SQLite database to your Chatbots.

Building a database helperWe’ll follow good coding practice and keep our database-related code separate from the main logic of our Bot.

Any code that touches the database will be confined to a new script, dbhelper.

py and we'll import this script into the main Bot script.

Modeling the problemWhenever you need to build a database, the first step is to think carefully about exactly how you’re going to represent the problem.

We’ll follow an iterative approach to building this bot, meaning that at each step, we’ll make the problem as simple as possible and then solve that before iterating on the project to build a more nuanced solution.

A To Do list consists of a number of items with text descriptions such as “Buy groceries”, or “Finish writing Chatbot tutorial”.

For now, we’ll model each item by simply storing its description in a table in a database.

We’ll see later on that this is a drastic oversimplification of the problem, but it’ll be good enough to get us started.

Our Chatbot will need to be able to add a new item to the database of items that it is remembering for the user, and it will also need to delete an item once the user has marked that item as done.

Furthermore, it’ll need to be able to get all the items to display them to the user.

Finally, it’ll need to be able to create the database table, in case we need to set up a new database (e.


when moving our bot to a new machine).

All of this functionality should be contained in our dbhelper.

py script, as all of it interacts with the database directly.

The dbhelper.

py codeCreate a new file called dbhelper.

py in the same directory as your Chatbot script, and add the following code:import sqlite3class DBHelper: def __init__(self, dbname="todo.

sqlite"): self.

dbname = dbname self.

conn = sqlite3.

connect(dbname) def setup(self): stmt = "CREATE TABLE IF NOT EXISTS items (description text)" self.


execute(stmt) self.


commit() def add_item(self, item_text): stmt = "INSERT INTO items (description) VALUES (?)" args = (item_text, ) self.


execute(stmt, args) self.


commit() def delete_item(self, item_text): stmt = "DELETE FROM items WHERE description = (?)" args = (item_text, ) self.


execute(stmt, args) self.


commit() def get_items(self): stmt = "SELECT description FROM items" return [x[0] for x in self.


execute(stmt)]In this code, we have five methods:__init__() takes a database name (by default store our data in a file called todo.

sqlite) and creates a database connection.

setup() creates a new table called items in our database.

This table has one column (called description)add_item() takes the text for the item and inserts it into our database table.

delete_item() takes the text for an item and removes it from the databaseget_items() returns a list of all the items in our database.

We use a list comprehension to take the first element of each item, as SQLite will always return data in a tuple format, even when there is only one column, so in this example each item we retrieve from the database will be similar to ("buy groceries", ) (a tuple) which the list comprehension converts to "buy groceries" (a simple string).

This DBHelper class that we've just created can be used by our Chatbot to do everything it needs to add, remove, and display items.

Let's modify the Chatbot code to use this new functionality.

Modifying the bot codeWe’ll be starting from where we left off in Part 1 of this tutorial, where we had a script called echobot.

py which echoed back every message that was sent to it.

To keep the different parts of this series separate, copy the echobot.

py to a new file called todobot.

py, and work with the new file instead.

The first change we need to make to the code is to import out DBHelper class and initialize an instance of it.

Add the following two lines near the top of the todobot.

py script:from dbhelper import DBHelperdb = DBHelper()With the code above, we can access the methods to get, add, and delete items through the db variable.

We previously wrote the function echo_all(), which looped through the updates and replied to each one of them.

We don't need this function now as we won't be echoing messages.

Delete it, and write a new function in its place called handle_updates()def handle_updates(updates): for update in updates["result"]: try: text = update["message"]["text"] chat = update["message"]["chat"]["id"] items = db.

get_items() if text in items: db.

delete_item(text) items = db.

get_items() else: db.

add_item(text) items = db.

get_items() message = ".".

join(items) send_message(message, chat) except KeyError: passThis is not too different from what we had before.

Again we loop through each update and grab the text and the chat components so that we can look at the text of the message we received and respond to the user who sent it.

Now, we also include a line to load all the items from our database and store them in the items variable.

We then check if the message that the user sent us is already one of the items we're storing.

If the user sends us a duplicate item, we'll take it as a signal to delete the item (i.


, mark it as done), but if the item is new, we'll add it to the database.

Either way, we'll send a message back to the user showing each of the saved items on a new line.

Asking our users to type out the items that they want to delete is not very user-friendly, but once we see that this is working, we’ll add some improvements.

Note that we’re already beginning to hit the limits of building Chatbots instead of traditional mobile apps: with Chatbots, all interactions have to be text-based, while with a traditional mobile app, we could add an “X” button on each item which the user could simply touch to remove that item.

We also need to call the setup() method of the database somewhere.

Although we only need to do this once, and it could be done in a separate script, we'll call it in the main function of our bot so that it gets called every time we start the bot.

Because if have the IF NOT EXISTS part in our CREATE TABLE statement, we'll still only create the table once.

Add the following line of code to the beginning of the main() function:db.

setup()Testing our first attempt at a To Do listTo make sure everything works as expected, run the todobot.

py script in your terminal and send a message to the Bot.

It should echo the message back to you.

This looks like what we had before, but there's a lot more going on under the hood.

Now any message you send gets stored in the database, and what looks like the echo is actually your entire To Do list, which contains only one item.

Send another (different) message, and you’ll see both messages be sent to you this time.

This is your updated To Do list.

To remove one of the items, send the text of the item that you want to remove.

Here we can see that we add “Buy eggs” to the list, and the entire list (consisting only of “Buy eggs”) gets returned.

Then we add “Finish writing”, and have both items returned.

When we send “Buy eggs” again, instead of adding a duplicate item, the bot deletes the existing one.

Adding a custom Telegram keyboardUX, or User eXperience is the magical term in application development that covers everything from the design of your interface to how responsive your application is to commands.

Basically, if users enjoy using your application and are able to control it intuitively, then your application has good UX.

At the moment, our UX is bad — users have to type out long item names to achieve the simple task of deleting the item, and to make it worse, we don’t even tell them that.

What we want, instead, is for our users to be able to select the items that they’re done with from a list, and have our bot delete the items that are tapped.

The way to achieve this in Telegram is by using a custom keyboard.

By default, when you chat to a user or a bot on Telegram, you are presented with your phone’s default keyboard layout (or nothing if you have a physical keyboard on your device).

As part of the sendMessage API call, we can pass along a custom keyboard, which will present the user with a keyboard that has specific keys with specific labels that we can specify.

When a user indicates that they want to mark items as done, we'll build a keyboard of all the existing items, and allow the user to press on the ones that they want to be deleted.

The syntax for building Telegram keyboards can look a bit complicated.

Each key is represented by a string, and each row on the keyboard is represented by a list of keys.

A keyboard is made up of a list of rows of keys, so if we wanted to rebuild a crude qwerty keyboard, we’d end up with a structure that looks like this: [ ["q", "w", "e", "…"], ["a", "s", "d", "…"], ["z", "x", "c"]].

Luckily the keyboard we want to construct is very simple — it should contain only one key per row, and that key should be one of the existing items on the list.

Add a build_keyboard() function to the todobot.

py which will take a list of items and construct a keyboard to allow the user to easily delete the items.

def build_keyboard(items): keyboard = [[item] for item in items] reply_markup = {"keyboard":keyboard, "one_time_keyboard": True} return json.

dumps(reply_markup)This function constructs the list of items, turning each item into a list to indicate that it should be an entire row of the keyboard.

We then build a dictionary which contains the keyboard as a value to the “keyboard” key and specifies that this keyboard should disappear once the user has made one choice (unlike with a normal keyboard where you might have to press many keys to construct a word, for this keyboard the user will only be choosing one item at a time).

Finally, we convert the Python dictionary to a JSON string, as this is the format that Telegram’s API expects from us.

We also need to teach our send_message() function how to include a custom keyboard when we want it to.

We'll add the keyboard as an optional parameter to this function, and if it's included we'll pass along the keyboard with the rest of the API call.

Change the send_message() function to look as follows:def send_message(text, chat_id, reply_markup=None): text = urllib.


quote_plus(text) url = URL + "sendMessage?text={}&chat_id={}&parse_mode=Markdown".

format(text, chat_id) if reply_markup: url += "&reply_markup={}".

format(reply_markup) get_url(url)Remember that the reply_markup argument that we pass along to Telegram isn't only the keyboard, but instead, an object that includes the keyboard along with other values, such as"one_time_keyboard": True.

Because we built the entire object in our build_keyboard() and encoded it as JSON, we can simply pass it along to Telegram in our send_message() function whenever necessary.

Choosing when to send the keyboardWe don’t always want to send the keyboard to the user.

By default, we want it to be as easy as possible for the user to add new items to the list.

So we’ll send the normal keyboard by default to allow the user to type out a new item to add to the list, and then require the user to input a special command in order to mark items as done.

Telegram commands by convention start with a forward slash, so we’ll ask our users to send the message /done when they want to check off items.

Although we could use any string and handle it as a command, following the convention and starting with a forward slash means that the user's Telegram client will autocomplete the command, or allow the user to click on the command from a previous message (which will be useful when we add the finishing touches to our Bot at the end of this post).

Now we’ll modify the handle_updates() function to send the custom keyboard to the user whenever they indicate that they want to mark items as done.

We'll also send the keyboard along after the user has just marked off an item, so several items can easily be deleted in a row.

The updated function should look as follows.

def handle_updates(updates): for update in updates["result"]: text = update["message"]["text"] chat = update["message"]["chat"]["id"] items = db.

get_items() if text == "/done": keyboard = build_keyboard(items) send_message("Select an item to delete", chat, keyboard) elif text in items: db.

delete_item(text) items = db.

get_items() keyboard = build_keyboard(items) send_message("Select an item to delete", chat, keyboard) else: db.

add_item(text) items = db.

get_items() message = ".".

join(items) send_message(message, chat)Try the bot out again.

Add some items and then send the /done message, and you should see a scrollable list of all the existing items presented that you can select to have them removed from the list, as in the image below.

Adding ownership of itemsOur To Do list is still crude, but it’s almost usable.

The last major feature that we have to add is the idea of ‘owners’ of different To Do lists.

Currently, our Bot only maintains a single To Do list in the database, so if multiple users were to use our bot then they would all share one giant list, and interfere with each other.

This would be entertaining (if you want some fun, try the bot out in its current form with some friends), but not very practical.

We want to modify the bot so that instead of returning every item in the database every time, it instead only returns the items associated with a specific chat_id.

In order to achieve this, we need to store the chat_id of the owner along with every item in the database.

Therefore, we need to add another column to our database, and our database logic and application logic to account for these changes.

Updating the dbhelper.

py fileThe first changes we’ll make are in the dbhelper.

py file.

We need to update:The setup() method to add the owner column when a new database is set upThe add_item() method to accept the chat_id (which we're using to identify owners) as an additional argument, and to add this to the database along with the itemThe delete_item() method, so that it only deletes items that match and that belong to the indicated ownerThe get_items() method to only return the items that belong to the specified ownerThe four updated methods for our DBHelper class can be seen below.

def setup(self): print("creating table") stmt = "CREATE TABLE IF NOT EXISTS items (description text, owner text)" self.


execute(stmt) self.


commit()def add_item(self, item_text, owner): stmt = "INSERT INTO items (description, owner) VALUES (?, ?)" args = (item_text, owner) self.


execute(stmt, args) self.


commit()def delete_item(self, item_text, owner): stmt = "DELETE FROM items WHERE description = (?) AND owner = (?)" args = (item_text, owner ) self.


execute(stmt, args) self.


commit()def get_items(self, owner): stmt = "SELECT description FROM items WHERE owner = (?)" args = (owner, ) return [x[0] for x in self.


execute(stmt, args)]Updating the todobot.

py fileThe changes here are simpler because of the abstraction provided by DBHelper.

The changes we have to make are in the handle_updates() function, and all we need to do is pass the chat_id along to DBHelper whenever we call one of the methods we just updated.

The new code for handle_updates() looks as follows:def handle_updates(updates): for update in updates["result"]: text = update["message"]["text"] chat = update["message"]["chat"]["id"] items = db.

get_items(chat) ## if text == "/done": keyboard = build_keyboard(items) send_message("Select an item to delete", chat, keyboard) elif text in items: db.

delete_item(text, chat) ## items = db.

get_items(chat) ## keyboard = build_keyboard(items) send_message("Select an item to delete", chat, keyboard) else: db.

add_item(text, chat) ## items = db.

get_items(chat) ## message = ".".

join(items) send_message(message, chat)The lines where we now have to pass chat along to the DBHelper have been indicated with ##.

The bot should now work with multiple users, assigning each user a new list.

Delete the todo.

sqlite file and start the bot again to recreate the database with the new changes.

Finishing touchesWe now have an MVP (minimum viable product) for our To Do bot.

To finish off, we’ll neaten up a couple of smaller issues.

One issue is that our bot interprets every message sent to it as an item, but in reality, users may send other messages, such as bot commands.

Another issue is that our database will become slower as it grows because we have no database indices.

Swallowing commands and adding a welcome textWhen a user starts our conversation with our bot for the first time, they have to send the /start command, and currently our bot adds this as the first item on their personal To Do list.

Our bot should ignore all messages that start with /, as these are Telegram commands.

We still want to recognize the /done command though, so we'll add two new logic blocks after catching the /done messages but before catching the items to handle / commands.

Add the following elif blocks to the handle_updates() function, directly before the elif text in items line.

elif text == "/start": send_message("Welcome to your personal To Do list.

Send any text to me and I'll store it as an item.

Send /done to remove items", chat)elif text.

startswith("/"): continueAdding database indexesChoosing where to put database indices is a classic space/time tradeoff in computer science.

If we add more indices, our database will often perform faster, but take drastically more storage space.

Because we’re dealing with very small data, we’ll add an index on both of our columns.

Adding the index to the owner column means that the get_items() method will be faster, as it will be quick to get all the items from a specific owner.

Adding an index to the items column means that the delete_item()method will run faster, as it won't have to search through all of a user's items to find one with matching text.

Modify the setup() method of DBHelper to look as follows:def setup(self): tblstmt = "CREATE TABLE IF NOT EXISTS items (description text, owner text)" itemidx = "CREATE INDEX IF NOT EXISTS itemIndex ON items (description ASC)" ownidx = "CREATE INDEX IF NOT EXISTS ownIndex ON items (owner ASC)" self.


execute(tblstmt) self.


execute(itemidx) self.


execute(ownidx) self.


commit()Again, we could simply run this once as a separate script, as the indices will persist, but it’s nice to have them saved in the setup() function in case we need to set up the bot on a new machine or create a new database for any reason.

The final code for the bot and the database helper can be found on GitHub.


. More details

Leave a Reply