view release on metacpan or  search on metacpan

lib/CGI/Info.pm  view on Meta::CPAN

		return $self->{is_ai};
	}

	# Allow environment variable override for testing or manual classification
	if(defined(my $override = $ENV{'IS_AI'})) {
		return $self->{is_ai} = $override ? 1 : 0;
	}

	my $agent = $ENV{'HTTP_USER_AGENT'};
	my $remote = $ENV{'REMOTE_ADDR'};

	unless($remote && $agent) {
		# Probably not running in CGI - assume not an AI crawler
		return 0;
	}

	# Known AI training and inference crawlers, matched against the User-Agent.
	# We intentionally do not consult the shared IP/agent cache here: is_robot()
	# stores 'robot' for many of the same UAs, and reading 'robot' != 'ai' would
	# produce a false negative.  Instance-level caching ($self->{is_ai}) above is
	# sufficient to avoid redundant regex evaluation within a single request.
	# Sources: vendor documentation and public bot lists.
	# Anthropic: ClaudeBot, Claude-Web, anthropic-ai
	# OpenAI: GPTBot, ChatGPT-User, OAI-SearchBot
	# Google: Google-Extended (AI training opt-out token)
	# Meta: meta-externalagent, FacebookBot (AI training)
	# Apple: Applebot-Extended (AI training subset)
	# Perplexity: PerplexityBot
	# Amazon: Amazonbot (Amazon AI / Alexa AI)
	# You.com: YouBot
	# Diffbot: Diffbot
	# Cohere: cohere-ai
	# Common Crawl: CCBot (primary data source for many LLM trainers)
	# ByteDance: Bytespider (TikTok / AI training)
	# Allen AI: AI2Bot
	# Timpi: TimpiBot
	if($agent =~ /ClaudeBot|Claude-Web|anthropic-ai|GPTBot|ChatGPT-User|OAI-SearchBot|Google-Extended|meta-externalagent|FacebookBot|Applebot-Extended|PerplexityBot|Amazonbot|YouBot|Diffbot|cohere-ai|CCBot|Bytespider|AI2Bot|TimpiBot/i) {
		# Enforce is_ai => is_robot so callers need not check both
		$self->{is_robot} = 1;
		return $self->{is_ai} = 1;
	}

	$self->{is_ai} = 0;
	return 0;
}

=head2 browser_type

Returns a string classifying the visitor's client.  The possible values are:

=over 4

=item * C<'mobile'> -- smartphone or tablet (checked first)

=item * C<'ai'> -- known AI training or inference crawler (see L</is_ai>)

=item * C<'search'> -- search-engine crawler

=item * C<'robot'> -- other automated client

=item * C<'web'> -- ordinary desktop or laptop browser

=back

    use Carp;
    use Template;
    use CGI::Info;

    my $info = CGI::Info->new();
    my $dir  = $info->rootdir() . '/templates/' . $info->browser_type();

    my $filename = ref($info);
    $filename =~ s/::/\//g;
    $filename = "$dir/$filename.tmpl";

    (-f $filename && -r $filename)
        or croak "Cannot open template '$filename'";

    my $template = Template->new();
    $template->process($filename, {}) or croak $template->error();

=cut

sub browser_type {
	my $self = shift;

	if($self->is_mobile()) {
		return 'mobile';
	}
	if($self->is_ai()) {
		return 'ai';
	}
	if($self->is_search_engine()) {
		return 'search';
	}
	if($self->is_robot()) {
		return 'robot';
	}
	return 'web';
}

=head2 get_cookie

Returns a cookie's value, or undef if no name is given, or the requested
cookie isn't in the jar.

Deprecated - use cookie() instead.

    use CGI::Info;

    my $i = CGI::Info->new();
    my $name = $i->get_cookie(cookie_name => 'name');
    print "Your name is $name\n";
    my $address = $i->get_cookie('address');
    print "Your address is $address\n";

=cut

sub get_cookie {
	my $self = shift;