This service will be shut down on 25th May 2018

Due to many reasons, we're going to shut down this service soon. You can use these alternatives:

• Chatbase — bot analytics from Google,
• Dashbot,
• Botanalytics,
• Amplitude — it has generous free plan with 10M events per month. It's a powerful analytics but have nothing specific to bots.

About this service

Botan is a telegram bot analytics system based on Yandex.Appmetrica. In this document you can find how to setup Yandex.Appmetrica account, as well as examples of Botan SDK usage.

Botan has 2 main use cases:

• Send to Botan info about every message sent by user and get basic usage stats like DAU, MAU, Retention, Commands and more complicated details.
• Get extended information about users by wrapping links you send to them — location, languages, devices and operating systems of your users.

Creating an account

• Register at http://appmetrica.yandex.com/
• After registration you will be prompted to create Application. Please use @YourBotName as a name.
• Save an API key from settings page, you will use it as a token for Botan API calls.
• Download lib for your language, and use it as described below. Don`t forget to insert your token!

SDK usage

We have libraries for the following languages:

• Go
• Java
• JavaScript
• PHP
• Python
• Ruby
• Rust
• Haskell

Alternatively, you can use Botan API via plain HTTP calls.

In case your preferred language is missed, you can make a contribution. It's easy — library usually contains 30 lines of code.

Also, pay attention to "what data to put into tracking data" section. 90% benefit from analytics usage lies in right integration;)

JavaScript example

Install npm: npm install botanio

varbotan=require('botanio')(token);botan.track(message,'Start');varuid=message.from.id;varurl='https://github.com/';// some url you want to send to userbotan.shortenUrl(uid,url,function(err,res,body){if(err){console.error(err);}else{console.log(body);// shortened url here}});

Python example

You need to install requests library to use python botan lib. You can do it with

pip install requests

Code:

importbotanbotan_token='.........'# Token got from @botaniobotuid=message.from_usermessage_dict=message.to_dict()
event_name=update.message.textprintbotan.track(botan_token, uid, message_dict, event_name)
.....
original_url= ... # some url you want to send to usershort_url=botan.shorten_url(original_url, botan_token, uid)
# now send short_url to user instead of original_url, and get geography, OS, Device of user

PHP example

You need to put the class in a convenient place.

private$token = 'token';
publicfunction_incomingMessage($message_json) {
$messageObj = json_decode($message_json, true);
$messageData = $messageObj['message'];
$botan = newBotan($this->token);
$botan->track($messageData, 'Start');
...$original_url = ...$uid = $message['from']['id']
$short_url = $botan->shortenUrl($url, $uid)
// now send short_url to user instead of original_url, and get geography, OS, Device of user
}

Ruby example

uid is a user id you get from Telegram.

require_relative 'botan'
token = 1111
uid = 1
message = { text: 'text' }
puts Botan.track(token, uid, message, 'Search')

Rust example

externcrate rustc_serialize;externcrate botanio;use botanio::{Botan};#[derive(Debug,RustcEncodable)]structMessage{some_metric:u32,another_metric:u32,}fnmain(){let token = "1111";let uid = 1;let name = "Search";let message = Message{some_metric:100,another_metric:500};let botan = Botan::new(token);
botan.track(uid, name,&message).unwrap();}

Java example

try (CloseableHttpAsyncClientclient = HttpAsyncClients.createDefault()) {
client.start();
Botanbotan = newBotan(client, newObjectMapper());
botan.track("1111", "1", ImmutableMap.of("some_metric": 100, "another_metric": 500), "Search").get();
}

Go example

package main
import (
"fmt""github.com/botanio/sdk/go"
)
typeMessagestruct {
SomeMetricintAnotherMetricint
}
funcmain() {
ch:=make(chanbool) // Channel for synchronizationbot:=botan.New("1111")
message:=Message{100, 500}
// Asynchronous track examplebot.TrackAsync(1, message, "Search", func(ans botan.Answer, err []error) {
fmt.Printf("Asynchonous: %+v\n", ans)
ch<-true// Synchronization send
})
// Synchronous track exampleans, _:=bot.Track(1, message, "Search")
fmt.Printf("Synchronous: %+v\n", ans)
<-ch// Synchronization receive
}

Haskell example

importNetwork.HTTP.Client (newManager)
importNetwork.HTTP.Client.TLS (tlsManagerSettings)
importServant.ClientimportWeb.Botan.SdkimportGHC.GenericsimportData.AesonimportControl.Concurrent.Asyncmain::IO()
main =do
manager <- runIO $ newManager tlsManagerSettings
let message = toJSON $AnyMessage"A""B"
a <- async $ track "token""user2222" message "test_action" manager
res <- wait a -- not real use caseprint res
dataAnyMessage=AnyMessage{a::Text
, b::Text}deriving (Show, Generic, ToJSON)

HTTP API

Track message

The base url is: https://api.botan.io/track

You should put data to Botan using POST method.

The url should look like https://api.botan.io/track?token=API_KEY&uid=UID&name=EVENT_NAME

Please provide a json document as the post body.

API response is a json document:

• on success: {"status": "accepted"}
• on failure: {"status": "failed"} or {"status": "bad request", "info": "some_additional_info_about_error"}

Shorten url

Send GET request to

https://api.botan.io/s/?token={token}&url={original_url}&user_ids={user_id}

You get shortened url in a plain-text response (in case the response code was 200). Codes other than 200 mean that an error occurred.

Also, in case of group chats you can add several user_ids: &user_ids={user_id_1},{user_id_2},{user_id_3}, but currently this data will not be used (because we don't know which particular user clicked link).

What to put into tracking data

###Basic integration

botan.track(<botan_token>, <user_who_wrote_to_bot>, <user_message_in_json_format>, <command_name>)

• command_name - we recommend to put here not just message text, but command. Example: user wrote '/search californication', put to command_name 'Search'. This will help you to aggregate type of user's input and get such report:
• user_message_in_json_format - whole message got from Telegram. For example, using python-telegram-bot you can do it in such way: message.to_dict(). Passing whole message, you will be able to see nice data like number of group chats among all chats: Also you will be able to get userids who performed some particular action (through segmentation) or your most active users and contact them:

###Advanced integration Actually, most benefit from analytics usage lies in sending right events with right data inside. Here is some best practices we recommend. Feel free to contribute your ways or improve existing ones.

#####Commands order That's how you can see what command users execute after which:

botan.track(<botan_token>, <user_who_wrote_to_bot>, {last_command: current_command}, "command_order")

Also you can send not pairs, but triples of commands:

botan.track(<botan_token>, <user_who_wrote_to_bot>, {before_last_command: {last_command: current_command}}, "command_order")

Using this, we can see, for example, what commands users execute after /start: #####Date cohorts Here is how you can tag every user with time cohort based on what was his first day at your service. Later you can use to see how your bot's performance has changed over time:

ifthis_is_first_occurence_of_user:
botan.track(<botan_token>,
<user_who_wrote_to_bot>,
{
'daily': message.date.strftime('%Y-%m-%d'),
'weekly': (message.date-datetime.timedelta(message.date.weekday())).strftime('%Y-%m-%d'),
'monthly': message.date.strftime('%Y-%m')
},
'cohorts')

Get user profiles by wrapping links

###How it works You create unique short link for each pair (user, link). When user clicks the link, Botan stores his user agent, IP address and other stuff. After that you'll be able to use user segmentation by geography, language, device and operating system (and see corresponding statistics).

###What url to wrap We suggest you to wrap every url that you send to user. Most often use case is sending "please rate us" link — most popular bots asks for rating in storebot.me.

###What you will get You'll get a lot of useful new data in the web interface:

Countries, regions and cities

Devices

Operating systems

Locales

###How to use Here you can find examples for Python, PHP. Feel free to make pull requests with wrappers for other languages (here's HTTP spec for the shortener).

##Contribution We are welcome any contributions as pull-requests!

Feel free to write more libraries for the languages we are not supporting yet.