What is your favorite memory of Bot-Depot?

Anonymous asks:

What is your favorite memory of Bot-Depot?

Oh, that's a name I haven't heard in a long time! I have many fond memories of Bot-Depot but one immediately jumped to mind which I'll write about below.

For some context for others: Bot-Depot was a forum site about chatbots that was most active somewhere in the range of the year 2002-08 or so, during what (I call) the "first wave of chatbots" (it was the first wave I lived through, anyway) - there was an active community of botmasters writing chatbots for the likes of AOL Instant Messenger, MSN Messenger, Yahoo! Messenger, and similar apps around that time (also ICQ, IRC, and anything else we could connect a bot to). The "second wave" was when Facebook Messenger, Microsoft Bot Platform, and so on came and left around the year 2016 or so.

My favorite memory about Bot-Depot was in how collaborative and innovative everyone was there: several of us had our own open source chatbot projects, which we'd release on the forum for others to download and use, and we'd learn from each other's code and make better and bigger chatbot programs. Some of my old chatbots have their code available at https://github.com/aichaos/graveyard, with the Juggernaut and Leviathan bots being some of my biggest that were written at the height of the Bot-Depot craze. Many of those programs aren't very useful anymore, since all the instant messengers they connected to no longer exist, and hence I put them up on a git repo named "graveyard" for "where chatbots from 2004 go to die" to archive their code but not forget these projects.

Most of the bots on Bot-Depot were written in Perl, and one particular chatbot I found interesting (and learned a "mind blowing" trick I could apply to my own bots) was a program called Andromeda written by Eric256, because it laid down a really cool pattern for how we could better collaborate on "plugins" or "commands" for our bots.

Many of the Bot-Depot bots would use some kind of general reply engine (like my RiveScript), and they'd also have "commands" like you could type /weather or /jokes in a message and it would run some custom bit of Perl code to do something useful separately from the regular reply engine. Before Andromeda gave us a better idea how to manage these, commands were a little tedious to manage: we'd often put a /help or /menu command in our bots, where we'd manually write a list of commands to let the users know what's available, and if we added a new command we'd have to update our /help command to mention it there.

Perl is a dynamic language that can import new Perl code at runtime, so we'd usually have a "commands" folder on disk, and the bot would look in that folder and require() everything in there when it starts up, so adding a new command was as easy as dropping a new .pl file in that folder; but if we forgot to update the /help command, users wouldn't know about the new command. Most of the time, when you write a Perl module that you expect to be imported, you would end the module with a line of code like this:

1;

And that's because: in Perl when you write a statement like require "./commands/help.pl"; Perl would load that code and expect the final statement of that code to be something truthy; if you forgot the "1;" at the end, Perl would throw an error saying it couldn't import the module because it didn't end in a truthy statement. So me and many others thought of the "1;" as just standard required boilerplate that you always needed when you want to import Perl modules into your program.

What the Andromeda bot showed us, though, is that you can use other 'truthy' objects in place of the "1;" and the calling program can get the value out of require(). So, Andromeda set down a pattern of having "self-documenting commands" where your command script might look something like:

# The command function itself, e.g. for a "/random 100" command that would
# generate a random number between 0 and 100.
sub random {
    my ($bot, $username, $message) = @_;
    my $result = int(rand($message));
    return "Your random number is: $result";
}

# Normally, the "1;" would go here so the script can be imported, but instead
# of the "1;" you could return a hash map that describes this command:
{
    command => "/random",
    usage => "/random [number]",
    example => "/random 100",
    description => "Pick a random number between 0 and the number you provide.",
    author => "Kirsle",
};

The chatbot program, then, when it imports your folder full of commands, it would collect these self-documenting objects from the require statements, like

# A hash map of commands to their descriptions
my %commands = ();

# Load all the command scripts from disk
foreach my $filename (<./commands/*.pl>) {
    my $info = require $filename;

    # Store their descriptions related to the command itself
    $commands{ $info->{'command'} } = $info;
}

And: now your /help or /menu command could be written to be dynamic, having it loop over all the loaded commands and automatically come up with the list of commands (with examples and descriptions) for the user. Then: to add a new command to your bot, all you do is drop the .pl file into the right folder and restart your bot and your /help command automatically tells a user about the new command!

For an example: in my Leviathan bot I had a "guess my number" game in the commands folder: https://github.com/aichaos/graveyard/blob/master/Leviathan/commands/games/guess.pl

Or a fortune cookie command: https://github.com/aichaos/graveyard/blob/master/Leviathan/commands/funstuff/fortune.pl

After we saw Andromeda set the pattern for self-documenting commands like this, I applied it to my own bots; members on Bot-Depot would pick one bot program or another that they liked, and then the community around that program would write their own commands and put them up for download and users could easily download and drop the .pl file into the right folder and easily add the command to their own bots!

I think there was some effort to make a common interface for commands so they could be shared between types of chatbot programs, too; but this level of collaboration and innovation on Bot-Depot is something I've rarely seen anywhere else since then.

We had also gone on to apply that pattern to instant messenger interfaces and chatbot brains, as well - so, Leviathan had a "brains" folder which followed a similar pattern: it came with a lot of options for A.I. engine to power your bot's general responses with, including Chatbot::Alpha (the precursor to my RiveScript), Chatbot::Eliza (an implementation of the classic 1970s ELIZA bot), and a handful of other odds and ends - designed in a "pluggable", self-documenting way where somebody could contribute a new type of brain for Leviathan and users could just drop a .pl file into the right folder and use it immediately. Some of our bots had similar interfaces for the instant messengers (AIM, MSN, YMSG, etc.) - so if somebody wanted to add something new and esoteric, like a CyanChat client, they could do so in a way that it was easily shareable with other botmasters.

For more nostalgic reading, a long time ago I wrote a blog post about SmarterChild and other AOL chatbots from around this time. I was meaning to follow up with an article about MSN Messenger but had never gotten around to it!