AI-Chat

 view release on metacpan or  search on metacpan

Changes  view on Meta::CPAN

Revision history for AI-Chat

0.6	18th August 2024
	 Changed default model from 'gpt-3.5-turbo-0125' to 'gpt-40-mini'
	 This model is cheaper, more capable and faster.

0.5     1st June 2024
         Added 'chat' method to support back and forth conversations
         Changed 'prompt' method to use the new 'chat' method

0.1     2nd March 2024
         First version, released on an unsuspecting world.

lib/AI/Chat.pm  view on Meta::CPAN

# Header for calling OpenAI
sub _get_header_openai {
    my $self = shift;
    $self->{'key'} = '' unless defined $self->{'key'};
    return {
         'Authorization' => 'Bearer ' . $self->{'key'},
         'Content-type'  => 'application/json'
     };
 }
 
 # Get a reply from a single prompt
 sub prompt {
     my ($self, $prompt, $temperature) = @_;
     
     $self->{'error'} = '';
     unless ($prompt) {
         $self->{'error'} = "Missing prompt calling 'prompt' method";
         return undef;
     }

    $temperature = 1.0 unless $temperature;

    my @messages;
    push @messages, {
        role    => 'system',
        content => $self->{'role'},
    } if $self->{'role'};
    push @messages, {
        role    => 'user',
        content => $prompt,
    };
    
    return $self->chat(\@messages, $temperature);
}

# Get a reply from a full chat
sub chat {
    my ($self, $chat, $temperature) = @_;
    
    if (ref($chat) ne 'ARRAY') {

lib/AI/Chat.pm  view on Meta::CPAN

=head1 SYNOPSIS

  use AI::Chat;

  my $chat  = AI::Chat->new(
      key   => 'your-api-key',
      api   => 'OpenAI',
      model => 'gpt-4o-mini',
  );

  my $reply = $chat->prompt("What is the meaning of life?");
  print $reply;

=head1 DESCRIPTION

This module provides a simple interface for interacting with AI Chat APIs,
currently supporting OpenAI.

The AI chat agent can be given a I<role> and then passed I<prompts>.  It will
reply to the prompts in natural language.  Being AI, the responses are
non-deterministic, that is, the same prompt will result in diferent responses
on different occasions.

Further control of the creativity of the responses is possible by specifying
at optional I<temperature> parameter.

=head1 API KEYS

A free OpenAI API can be obtained from L<https://platform.openai.com/account/api-keys>

=head1 MODELS

Although the API Key is free, each use incurs a cost.  This is dependent on the
number of tokens in the prompt and the reply.  Different models have different costs.
The default model C<gpt-4o-mini> is the lowest cost of the useful models and
is a good place to start using this module.  Previous versions of this module
defaulted to C<gpt-3.5-turbo-0125> but the current default is cheaper and
quicker. For most purposes, the default model should be used.

See also L<https://platform.openai.com/docs/models/overview>

=head1 METHODS

=head2 new

lib/AI/Chat.pm  view on Meta::CPAN

=item model

The language model to use (default: 'gpt-4o-mini').

See L<https://platform.openai.com/docs/models/overview>

=item role

The role to use for the bot in conversations.

This tells the bot what it's purpose when answering prompts.

For example: "You are a world class copywriter famed for
creating content that is immediately engaging with a
lighthearted, storytelling style".

=item debug

Used for testing.  If set to any true value, the prompt method
will return details of the error encountered instead of C<undef>

=back

=head2 prompt

  my $reply = $chat->prompt($prompt, $temperature);

Sends a prompt to the AI Chat API and returns the response.

=head3 Parameters

=over 4

=item prompt

C<required> The prompt to send to the AI.

This is a shorthand for C<chat> when only a single response is needed.

=item temperature

The creativity level of the response (default: 1.0).

Temperature ranges from 0 to 2.  The higher the temperature,
the more creative the bot will be in it's responses.

=back

=head2 chat

  my $reply = $chat->prompt(\@chat, $temperature);

Sends a multi-message chat to the AI Chat API and returns the response.

Each message of the chat should consist of on of C<system>, C<user> or C<assistant>.
Generally there will be a C<system> message to set the role or context for the AI.
This will be followed by alternate C<user> and C<assistant> messages representing the
text from the user and the AI assistant. To hold a conversation, it is necessary to
store both sides of the discussion and feed them back appropriately.

  my @chat;