Telegram library for the Elixir language.
It provides:
- an inteface to the Telegram Bot HTTP-based APIs (
Telegram.Api
) - a couple of bot behaviours to define you bots (
Telegram.Bot
,Telegram.ChatBot
) - two bot runners (
Telegram.Poller
,Telegram.Webhook
)
The package can be installed by adding telegram
to your list of dependencies in mix.exs
:
def deps do
[
{:telegram, github: "visciang/telegram", tag: "xxx"}
]
end
This module expose a light layer over the Telegram Bot API HTTP-based interface, it does not expose any "(data)binding" over the HTTP interface and tries to abstract away only the boilerplate for building / sending / serializing the API requests.
Compared to a full data-binded interface it could result less "typed frendly" but it will work with any version of the Bot API, hopefully without updates or incompatibily with new Bot API versions (as much as they remain backward compatible).
References:
Given the token of your Bot you can issue any request using:
- method: Telegram API method name (ex. "getMe", "sendMessage")
- options: Telegram API method specific parameters (you can use Elixir's native types)
Given the bot token (something like):
token = "123456:ABC-DEF1234ghIkl-zyx57W2v1u123ew11"
Telegram.Api.request(token, "getMe")
{:ok, %{"first_name" => "Abc", "id" => 1234567, "is_bot" => true, "username" => "ABC"}}
Telegram.Api.request(token, "sendMessage", chat_id: 876532, text: "Hello! .. silently", disable_notification: true)
{:ok,
%{"chat" => %{"first_name" => "Firstname",
"id" => 208255328,
"last_name" => "Lastname",
"type" => "private",
"username" => "xxxx"},
"date" => 1505118722,
"from" => %{"first_name" => "Yyy",
"id" => 234027650,
"is_bot" => true,
"username" => "yyy"},
"message_id" => 1402,
"text" => "Hello! .. silently"}}
Telegram.Api.request(token, "getUpdates", offset: -1, timeout: 30)
{:ok,
[%{"message" => %{"chat" => %{"first_name" => "Firstname",
"id" => 208255328,
"last_name" => "Lastname",
"type" => "private",
"username" => "xxxx"},
"date" => 1505118098,
"from" => %{"first_name" => "Firstname",
"id" => 208255328,
"is_bot" => false,
"language_code" => "en-IT",
"last_name" => "Lastname",
"username" => "xxxx"},
"message_id" => 1401,
"text" => "Hello!"},
"update_id" => 129745295}]}
If an API parameter has a InputFile
type and you want to send a local file,
for example a photo stored at "/tmp/photo.jpg", just wrap the parameter
value in a {:file, "/tmp/photo.jpg"}
tuple. If the file content is in memory
wrap it in a {:file_content, data, "photo.jpg"}
tuple.
Telegram.Api.request(token, "sendPhoto", chat_id: 876532, photo: {:file, "/tmp/photo.jpg"})
Telegram.Api.request(token, "sendPhoto", chat_id: 876532, photo: {:file_content, photo, "photo.jpg"})
To download a file from the telegram server you need a file_path
pointer to the file.
With that you can download the file via Telegram.Api.file
.
{:ok, res} = Telegram.Api.request(token, "sendPhoto", chat_id: 12345, photo: {:file, "example/photo.jpg"})
# pick the 'file_obj' with the desired resolution
[file_obj | _] = res["photo"]
# get the 'file_id'
file_id = file_obj["file_id"]
{:ok, %{"file_path" => file_path}} = Telegram.Api.request(token, "getFile", file_id: file_id)
{:ok, file} = Telegram.Api.file(token, file_path)
If an API parameter has a non primitive scalar type it is explicitly pointed out as "A JSON-serialized object"
(ie InlineKeyboardMarkup
, ReplyKeyboardMarkup
, etc).
In this case you can wrap the parameter value in a {:json, value}
tuple.
sendMessage with keyboard
keyboard = [
["A0", "A1"],
["B0", "B1", "B2"]
]
keyboard_markup = %{one_time_keyboard: true, keyboard: keyboard}
Telegram.Api.request(token, "sendMessage", chat_id: 876532, text: "Here a keyboard!", reply_markup: {:json, keyboard_markup})
Check the examples under example/example_*.exs
.
You can run them as a Mix
self-contained script.
BOT_TOKEN="..." example/example_chatbot.exs
The Telegram platform supports two ways of processing bot updates, getUpdates
and setWebhook
.
getUpdates
is a pull mechanism, setWebhook
is a push mechanism. (ref: bots webhook)
This library currently implements both models via two supervisors.
This mode can be used in a dev environment or if your bot doesn't need to "scale". Being in pull it works well behind a firewall (or behind a home internet router).
Refer to the Telegram.Poller
module docs for more info.
The Telegram HTTP Client is based on Tesla
.
The Tesla.Adapter
and options should be configured via the [:tesla, :adapter]
application environment key.
(ref. https://hexdocs.pm/tesla/readme.html#adapters)
For example, a good default could be:
config :tesla, adapter: {Tesla.Adapter.Hackney, [recv_timeout: 40_000]}
a dependency should be added accordingly in your mix.exs
:
defp deps do
[
{:telegram, github: "visciang/telegram", tag: "xxx"},
{:hackney, "~> 1.18"},
# ...
]
end
This mode interfaces with the Telegram servers via a webhook, best for production use.
The app is meant to be served over HTTP, a reverse proxy should be placed in front of it, facing the public network over HTTPS.
It's possible to use two Plug
compatible webserver: Bandit
and Plug.Cowboy
.
Alternatively, if you have a Phoenix
/ Plug
based application facing internet, you can directly integrate the webhook.
Refer to the Telegram.Webhook
module docs for more info.
We can define stateless / stateful bot.
-
A stateless Bot has no memory of previous conversations, it just receives updates, process them and so on.
-
A stateful Bot instead can remember what happened in the past. The state here refer to a specific chat, a conversation (chat_id) between a user and a bot "instance".
Telegram.Bot
: works with the stateless async dispatch modelTelegram.ChatBot
: works with the stateful chat dispatch model
The library attaches two metadata fields to the internal logs: [:bot, :chat_id]. If your app runs more that one bot these fields can be included in your logs (ref. to the Logger config) to clearly identify and "trace" every bot's message flow.
A chat_bot app, deployed to Gigalixir PaaS and served in webhook mode: https://github.com/visciang/telegram_example