head

root

para

L<Examples| file:../examples/index.podlite>


L<Download |file:../download/index.podlite>


React

=begin React :component<HeaderCol> :id<menu> 
=para L<Documentation|/doc/introduction>
=para L<Modules| /mods>
=para L<Examples| file:../examples/index.podlite>
=para L<Download |file:../download/index.podlite>
=para L<About|file:../about/index.podlite>
=para 🔍 K<⌘K>/K<ctrl-K>
=end React


HEADER

FOOTER

README

unit module LLM::DWIM;
use LLM::Functions;
use LLM::Prompts;
use JSON::Fast;
use Log::Async;
use TOML;

logger.untapped-ok = True;

state &evaluator;

my $default-evaluator = 'OpenAI';

sub dwim(Str $str) is export {
  my $base = %*ENV<XDG_HOME> // $*HOME.child('.config');
  my $conf-file = %*ENV<DWIM_LLM_CONF> // $base.child('llm-dwim.toml');
  my $conf = do {
    if $conf-file.IO.e {
      debug "Reading configuration from $conf-file";
      from-toml($conf-file.IO.slurp);
    } else {
      debug "No configuration file found at $conf-file, using default: $default-evaluator";
      %( evaluator => $default-evaluator );
    }
  }
  my $evaluator = $conf<evaluator> or die "No evaluator in $conf-file (found: { $conf.keys })";
  my $evaluator-config = $conf{ $evaluator } // {};
  debug "Configuration: $evaluator, { $evaluator-config.raku }";
  &evaluator //= llm-function(
    llm-evaluator => llm-configuration( $evaluator, |%( $evaluator-config ) )
  );
  my $msg = llm-prompt-expand($str);
  debug "sending $msg";
  evaluator($msg);
}



LLM::DWIM -- Do What I Mean, with help from large language models.


code

=begin code

use LLM::DWIM;

say dwim "How many miles is it from the earth to the moon?";
# Approximately 238,900 miles (384,400 kilometers)

say dwim "@NothingElse How many miles is it from the earth to the moon? #NumericOnly";
# 238900

sub distance-between($from,$to) {
  dwim "@NothingElse #NumericOnly What is the distance in miles between $from and $to?";
}

say distance-between("earth","sun");
# 92955887.6 miles

=end code


=begin code

evaluator = "gemini"
gemini.temperature = 0.5

=end code


This is a simple wrapper around L<LLM::Functions|https://raku.land/zef:antononcube/LLM::Functions>,
and L<LLM::Prompts|https://raku.land/zef:antononcube/LLM::Prompts>
It provides a single subroutine, C<dwim>, that sends a string to an LLM evaluator, making use of
a configuration file to say a little more about what you mean.


This function takes a string, expands it using L<LLM::Prompts>, and uses L<LLM::Functions> to
evaluate the string.


=begin code

use LLM::Functions;
use LLM::Prompts;
use TOML;

my $conf-dir = %*ENV<XDG_HOME> // $*HOME.child('.config');
my $conf = from-toml($conf-dir.child('llm-dwim.toml').IO.slurp);
my $evaluator = $conf<evaluator>;
my &evaluator //= llm-function(
    llm-evaluator => llm-configuration( $evaluator, |%( $conf{ $evaluator } ) )
);
my $msg = llm-prompt-expand($str);
evaluator($msg);

=end code


For diagnostics, use L<Log::Async|https://raku.land/cpan:BDUGGAN/Log::Async> and add a tap, like so:


=begin code

use LLM::DWIM;
use Log::Async;

logger.send-to($*ERR);

say dwim "How many miles is it from earth is the moon? #NumericOnly";

=end code


This module looks for C<llm-dwim.toml> in either C<XDG_HOME> or C<HOME/.config>.
This can be overridden by setting C<DWIM_LLM_CONF> to another filename.


The configuration file should be in TOML format and should contain at least one key,
C<evaluator>, which should be the name of the LLM evaluator to use.  Evaluators
can be configured using TOML syntax, with the evaluator name as the key.


=begin code

evaluator = "gemini"

=end code


=begin code

evaluator = "OpenAI"
OpenAI.temperature = 0.9
OpenAI.max-tokens = 100

=end code


See L<LLM::Functions|https://raku.land/zef:antononcube/LLM::Functions> for all of
the configuration options.


Also, this package includes a `llm-dwim` script that can be used
to evaluate a string from the command line. It will look for the
configuration file in the same way as the module.  It accepts
either a string or a list of words.  The single argument "-" will
cause it to read from standard input.


llm-dwim -h  # get usage
llm-dwim "How many miles is it from the earth to the moon?"
llm-dwim -v how far is it from the earth to the moon\?
echo "what is the airspeed velocity of an unladen swallow?" | llm-dwim -


L<LLM::Functions|https://raku.land/zef:antononcube/LLM::Functions>,
L<LLM::Prompts|https://raku.land/zef:antononcube/LLM::Prompts>


This was inspired by the also excellent L<DWIM::Block|https://metacpan.org/pod/DWIM::Block> module.


DWIM

How-to-use

The Raku Knowledge Base


=begin React :component<HeaderCol> :id<footer>  
=begin nested :!nested
=item1 B<Language>
    =item2 L<Get started|file:../getting-started/getting-started.podlite>
    =item2 L<Why Raku?|/doc/language/faq#Why-should-I-learn-Raku-What's-so-great-about-it>
    =item2 L<Try Raku|https://glot.io/new/raku>
    =item2 L<Raku cheat sheet|https://github.com/Raku/mu/blob/master/docs/Perl6/Cheatsheet/cheatsheet.txt>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<Raku on Exercism|https://exercism.org/tracks/raku>
    =item2 L<Wikipedia|https://en.wikipedia.org/wiki/Raku_(programming_language)>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>

=end nested
=begin nested :!nested

=item1 B<L<Documentation|/doc>>
    =item2 L<Getting started, Migration guides from other languages, & Tutorials | file:../doc/introduction.podlite>
    =item2 L<Language References|file:../doc/reference.podlite>
    =item2 L<Type Reference| file:../doc/types.podlite>
    =item2 L<Miscellaneous| file:../doc/miscellaneous.podlite>
    =item2 L<FAQs (Frequently Asked Questions)|/doc/language/faq>
    =item2 L<Community|/doc/language/community>
    =item2 L<The list of all documents|file:../doc/index.podlite>

=item1 B<Resources>
    =item2 L<The Raku Guide|https://raku.guide/>
    =item2 L<Books |https://perl6book.com/>
    =item2 L<Rosetta Code | https://www.raku.org/community/rosettacode>

=end nested
=begin nested :!nested

=item1 B<Resoures>
    =item2 L<Download | file:../download/index.podlite>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>

=item1 B<Learning>
    =item2 L<The Raku Guide|https://raku.guide/>
    =item2 L<Wikibook |https://en.wikibooks.org/wiki/Raku_Programming>
    =item2 L<Books |https://perl6book.com/>
    =item2 L<Rosetta Code | https://www.raku.org/community/rosettacode>
    =item2 L<Learn Raku in Y minutes|https://learnxinyminutes.com/docs/raku/>
    =item2 L<Golfing |https://github.com/AlexDaniel/raku-golf-cheatsheet>
=end nested
=begin nested :!nested

=item1 B<Explore>
    =item2 L<Raku Blog Aggregator|https://planet.raku.org/>
    =item2 L<Rakudo Weekly|https://rakudoweekly.blog/>
    =item2 L<The Weekly Challenge |https://perlweeklychallenge.org/>
    =item2 L<Raku Advent Calendar|https://raku-advent.blog/>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>
    =item2 L<empty|#>

=end nested
=end React


=begin React :component<CookieConsent> :id<CookieConsent> :buttonCaption("Got it!")
=para
This website uses cookies for analytics.
=end React


README

Raku-LLM-Containerization

LLM::Containerization v0.1.0

Authors

License

Dependencies

Test Dependencies

Provides

Documentation