A plugins based bot for Telegram messaging app written in Ruby.
Install required dependencies:
$ sudo apt-get install ffmpeg wget
Clone this repository and then run bundle
to install the dependencies.
$ git clone git@github.com:syxanash/joshua_bot.git
$ cd joshua_bot
$ bundle install
Enter a valid bot API token in config.json
(see Configure below) then run the bot with:
$ ruby bot.rb
A log file will be created in /tmp/joshua_bot_tmp
directory with a random name (e.g. bot_1b3510f004bd.log
)
Before running the bot you need to add a telegram API key for your bot to the file config.json
. Check out the official telgram documentation to achieve this.
Here's an example of a config.json
once you have your API key set up:
{
"token": "103414657:AAGh0I6l-CKf_TDu6CUNa7c7MgnfRbUDzMQ",
"prod": true,
"temp_directory": "/tmp/joshua_bot_tmp",
"plugin_folder": "basic",
"password": "",
"startup_commands": ["/xkcd"],
"openai": {
"token": "",
"recognize_plugins": true,
"max_interaction_history": 5,
"log_prompts": false
}
}
If you take a look at the folder lib/plugins/
you will find different sub folders containing various plugins inside. By default the plugins loaded on the bot are in basic/
. Alternatively you can specify a different folder in the config.json
modifying the value plugin_folder
(e.g. "spioncino"
) .
You can create your own folder of plugins and drag plugins from other folders by making a new folder in lib/plugins/
, for instance:
$ mkdir lib/plugins/workinprogress
Once you've done that you can specify the folder "workinprogress"
inside config.json
.
Notice that the plugin help.rb
will always be loaded since it resides in the root directory of the plugins lib/plugins
. You can place in this folder all the plugins which will always be loaded by the bot.
If you want to be the only one talking to your bot you can add a tiny layer of security setting up a bot password. By doing so everytime you start the conversation with the bot it will ask you for a password before interpreting various commands.
You can set a bot password in the config.json
under the value "password"
.
Each plugin is a class which extends the class AbsPlugin
. A plugin should have the following four methods:
initialize
(not mandatory)command
show_usage
do_stuff
examples
(see OpenAI Integration)
You can place a new plugin in the folder lib/plugins
or inside a subfolder of this directory to group plugins all your plugins. For example basic/
and spioncino/
are two different plugin folders.
It's a method used to initialize class variables and other settings used by the plugin. This method will be called when the bot is loading the plugins for the first time. Also see Ruby object initialization.
This method should return the regular expression used to match a command which the bot will interpret. For instance:
def command
/^\/fortune$/
end
In this case the plugin will be activated upon entering the command /fortune
. A slightly more complex command could be instad:
def command
/(^\/takephoto$|photo)/
end
Which will interpret either /takephoto
or simply photo
.
You can display a help message to show the user the correct usage of the plugin. This method will be called when the user invokes the command in a wrong way for example if the command requires additional parameters.
For instance the plugin Spione will call the method show_usage
if the user invokes the command /spione
without parameters. A correct usage of the plugin Spione would be /spione on
.
This method will contain the actual code executed when the command is invoked. You can send simple text messages to the user (see the plugin Fortune), images (see Xkcd plugin) or audio messages (see Say).
You can interact with plugins either by passing additional parameters when calling the plugin or by asking the user further questions by reading the buffer. An analogy to this could be passing information to a command line tool either by passing arguments or using STDIN
😄
Additional command's parameters will be stored in the formal parameter match_result
of the method do_stuff
. The plugin's additional parameters will be created if you define a regular expression which accepts extra parameters like:
/^\/diceroll\s?(\d*?)?$/
In this case DiceRoll plugin can be invoked by using the command /diceroll
or you can pass an extra parameter as an integer like so: /diceroll 50
(this plugin will now generate a random number from 1-50). match_result
is an array which will contain all the captured variables of the matched plugin's regex.
If you want your plugins to be more interactive asking questions and evaluating replies, more like a conversation with the user, you can use read_buffer()
. This method must be invoked in the do_stuff
block of your plugin.
Here's an example of a simple plugin which makes the addition of two numbers:
def do_stuff(match_results)
bot.api.send_message(chat_id: message.chat.id, text: "what's the first number?")
first_number = read_buffer
bot.api.send_message(chat_id: message.chat.id, text: "what's the second number?")
second_number = read_buffer
bot.api.send_message(chat_id: message.chat.id, text: "I'm not a calculator so F𝔘Ⓒk off")
sleep(3)
bot.api.send_message(chat_id: message.chat.id, text: "it's #{first_number.to_i + second_number.to_i}...")
end
And this is going to be the result:
Some plugins which use the session buffer are:
When a bot is protected by password you can execute a sequence of commands as soon as the user logs in. Take a look at config.json
you'll find an array called startup_commands
. You can edit this variable and add your list of startup commands that you wish to execute after logging in.
For example:
"startup_commands": ["/diceroll 12", "/ping", "/morra"]
And this will be the output after entering the password correctly:
In order to stop a bot from asking further inputs and replies from a user, perhaps because of a bug or an unforeseen loop condition in your code, you can type the command /cancel
and the plugin will terminate its execution.
By default, a log file will be created in /tmp/joshua_bot_tmp
(you can change this by editing the temp_directory
name in config.json
). Alternatively, if you'd like to redirect all output to STDOUT while testing your bot, simply change the variable prod
to false
in config.json
.
Thanks to the OpenAI APIs, you can chat with Joshua Bot using ChatGPT. This feature is completely optional and this bot will work even without OpenAI. OpenAI APIs are used to interpret commands of various plugins or to just talk to your bot using ChatGPT.
To enable the bot to use OpenAI, enter your OpenAI API Token in config.json
.
This feature is toggled on by default, so whenever you send a message to the bot without directly invoking a command, it will automatically try to interpret it as a command for executing a plugin.
(In this picture the XKCD comic, the dice roll number and battery status are plugins executed by the bot)
In order to let the bot recognize and execute plugins simply by chatting with it, you will need to add the examples when creating new custom plugins. Examples will be used to train the bot to recognize the text based on the description and execute the related command.
This is the format you will need to follow:
def examples
[
{ command: '/diceroll', description: 'return a random number from 1 to 6' },
{ command: '/diceroll 10', description: 'return a random number from 1 to 10' }
]
end
Make sure the command in examples matches the regex in command
method of your plugin!
Take a look at the demo plugins to have an idea of how to write your own plugin and to make the bot execute commands: morra.rb, lyrics.rb, randomlogo.rb
If you'd like to talk to the bot only using ChatGPT without any command translation, you can toggle this feature off by setting the flag recognize_plugins
to false
in config.json
.
The default personality of Joshua is the following:
Joshua is a helpful chatbot who enjoys chatting with any human who interacts with him.
This was hardcoded inside ai_handler.rb. You can add extra traits to your bot in a personality file, which by default is called personality.txt
. You can also rename or specify your own file in config.json
.
Here's another example of more personality you can add to your bot inside personality.txt
:
He's a very sarcastic bot that lives inside a Raspberry Pi (Zero W to be precise).
He really likes the movie WarGames and always tries his best to put some quotes from this movie.
When OpenAI integration is enabled and the user sends a voice message to the bot, this will automatically be transcribed using Whisper APIs. The transcription will then be used for interpreting plugins or communicating with ChatGPT.
You should definitely tell your partner about an awesome implementation of this bot as a simple surveillance system for your mansion: spioncino