teloxide

A full-featured framework that empowers you to easily build Telegram bots using the async/.await syntax in Rust. It handles all the difficult stuff so you can focus only on your business logic.

Table of contents

• Features
• Getting started
• Examples
  • The ping-pong bot
  • Commands
  • Guess a number
• More examples!
• Recommendations
• FAQ
  • Where I can ask questions?
  • Why Rust?
• Community bots
• Contributing

Features

• Type-safe. teloxide leverages the Rust's type system with two serious implications: resistance to human mistakes and tight integration with IDEs. Write fast, avoid debugging as much as possible.

• Flexible API. teloxide gives you the power of streams: you can combine all 30+ patterns when working with updates from Telegram.

• Persistency. By default, teloxide stores all user dialogues in RAM, but you can store them somewhere else (for example, in DB) just by implementing 2 functions.

• Convenient dialogues system. Define a type-safe finite automaton and transition functions to drive a user dialogue with ease (see the guess-a-number example below).

Getting started

1. Create a new bot using @Botfather to get a token in the format 123456789:blablabla.
2. Initialise the TELOXIDE_TOKEN environmental variable to your token:

# Unix
$ export TELOXIDE_TOKEN=<Your token here>

# Windows
$ set TELOXIDE_TOKEN=<Your token here>

3. Be sure that you are up to date:

$ rustup update stable

4. Execute cargo new my_bot, enter the directory and put these lines into your Cargo.toml:

[dependencies]
teloxide = "0.1.0"
log = "0.4.8"
tokio = "0.2.11"
pretty_env_logger = "0.4.0"

The ping-pong bot

This bot has a single message handler, which answers "pong" to each incoming message:

(Full)

use teloxide::prelude::*;

#[tokio::main]
async fn main() {
    teloxide::enable_logging!();
    log::info!("Starting ping_pong_bot!");

    let bot = Bot::from_env();

    Dispatcher::new(bot)
        .messages_handler(|rx: DispatcherHandlerRx<Message>| {
            rx.for_each(|message| async move {
                message.answer("pong").send().await.log_on_error().await;
            })
        })
        .dispatch()
        .await;
}

Commands

Commands are defined similar to how we define CLI using structopt. This bot says "I am a cat! Meow!" on /meow, generates a random number within [0; 1) on /generate, and shows the usage guide on /help:

(Full)

// Imports are omitted...

#[derive(BotCommand)]
#[command(rename = "lowercase", description = "These commands are supported:")]
enum Command {
    #[command(description = "display this text.")]
    Help,
    #[command(description = "be a cat.")]
    Meow,
    #[command(description = "generate a random number within [0; 1).")]
    Generate,
}

fn generate() -> String {
    thread_rng().gen_range(0.0, 1.0).to_string()
}

async fn answer(
    cx: DispatcherHandlerCx<Message>,
    command: Command,
) -> ResponseResult<()> {
    match command {
        Command::Help => cx.answer(Command::descriptions()).send().await?,
        Command::Generate => cx.answer(generate()).send().await?,
        Command::Meow => cx.answer("I am a cat! Meow!").send().await?,
    };

    Ok(())
}

async fn handle_commands(rx: DispatcherHandlerRx<Message>) {
    // Only iterate through commands in a proper format:
    rx.commands::<Command>()
        // Execute all incoming commands concurrently:
        .for_each_concurrent(None, |(cx, command, _)| async move {
            answer(cx, command).await.log_on_error().await;
        })
        .await;
}

#[tokio::main]
async fn main() {
    // Setup is omitted...
}

See? The dispatcher gives us a stream of messages, so we can handle it as we want! Here we use our .commands::<Command>() and .for_each_concurrent(), but others are also available:

• .flatten()
• .left_stream()
• .scan()
• .skip_while()
• .zip()
• .select_next_some()
• .fold()
• .inspect()
• ... And lots of others!

Guess a number

Wanna see more? This is a bot, which starts a game on each incoming message. You must guess a number from 1 to 10 (inclusively):

(Full)

// Setup is omitted...

#[derive(SmartDefault)]
enum Dialogue {
    #[default]
    Start,
    ReceiveAttempt(u8),
}

async fn handle_message(
    cx: DialogueDispatcherHandlerCx<Message, Dialogue>,
) -> ResponseResult<DialogueStage<Dialogue>> {
    match cx.dialogue {
        Dialogue::Start => {
            cx.answer(
                "Let's play a game! Guess a number from 1 to 10 (inclusively).",
            )
            .send()
            .await?;
            next(Dialogue::ReceiveAttempt(thread_rng().gen_range(1, 11)))
        }
        Dialogue::ReceiveAttempt(secret) => match cx.update.text() {
            None => {
                cx.answer("Oh, please, send me a text message!").send().await?;
                next(cx.dialogue)
            }
            Some(text) => match text.parse::<u8>() {
                Ok(attempt) => match attempt {
                    x if !(1..=10).contains(&x) => {
                        cx.answer(
                            "Oh, please, send me a number in the range [1; \
                             10]!",
                        )
                        .send()
                        .await?;
                        next(cx.dialogue)
                    }
                    x if x == secret => {
                        cx.answer("Congratulations! You won!").send().await?;
                        exit()
                    }
                    _ => {
                        cx.answer("No.").send().await?;
                        next(cx.dialogue)
                    }
                },
                Err(_) => {
                    cx.answer(
                        "Oh, please, send me a number in the range [1; 10]!",
                    )
                    .send()
                    .await?;
                    next(cx.dialogue)
                }
            },
        },
    }
}

#[tokio::main]
async fn main() {
    // Setup is omitted...
}

Our finite automaton, designating a user dialogue, cannot be in an invalid state, and this is why it is called "type-safe". We could use enum + Options instead, but it will lead is to lots of unpleasure .unwrap()s.

Remember that a classical finite automaton is defined by its initial state, a list of its possible states and a transition function? We can think that Dialogue is a finite automaton with a context type at each state (Dialogue::Start has (), Dialogue::ReceiveAttempt has u8).

See examples/dialogue_bot to see a bit more complicated bot with dialogues.

More examples!

Recommendations

• Use this pattern:

#[tokio::main]
async fn main() {
    run().await;
}

async fn run() {
    // Your logic here...
}

Instead of this:

#[tokio::main]
async fn main() {
    // Your logic here...
}

The second one produces very strange compiler messages because of the #[tokio::main] macro. However, the examples in this README use the second variant for brevity.

FAQ

Where I can ask questions?

Issues is a good place for well-formed questions, for example, about the library design, enhancements, bug reports. But if you can't compile your bot due to compilation errors and need quick help, feel free to ask in our official group: https://t.me/teloxide.

Why Rust?

Most programming languages have their own implementations of Telegram bots frameworks, so why not Rust? We think Rust provides enough good ecosystem and the language itself to be suitable for writing bots.

Community bots

Feel free to push your own bot into our collection: https://github.com/teloxide/community-bots. Later you will be able to play with them right in our official chat: https://t.me/teloxide (coming soon...).

Contributing

See CONRIBUTING.md.