tiktoken_ruby 0.0.2-x86_64-darwin → 0.0.3-x86_64-darwin

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: bd7631d7d4eb434ad4f3a55334c4827ad81131b80b0ce387edf43ad38ccc14d9
-  data.tar.gz: b097879f6f4118ab9cb4f4d651c6f0ab9915b4f035276fc86e008101884bbc85
+  metadata.gz: 5ff2e729499103996a4b72baf94f7130746ca7ce453597c492ed1a679b5faab3
+  data.tar.gz: 583ff111ac69257887fdfae23c84420b9f305ac36ba37450fbbc87e4e5c707dd
 SHA512:
-  metadata.gz: ee0a642763f4bbbb641abe64db37e8e3a263a8fe9b64ccbbd673c2121667145b1c57b3f03b73c9bf583019156f2d0b7577592a7ace59a782d86bb6d419b062ba
-  data.tar.gz: 67f87705d9019b8c156fbe38e8fd9e947f8d555df69e6bec0a1c520f426cfa92e95878e234c91e344fe956765b5f80d63661f298db01985234f4220291596aba
+  metadata.gz: 5b9961606f15107c39547687741c39631da1244d9d44ac7d78347ebc13a6088aa0e13451779a7f4aa064ec9b929231c6777ceac296f730a8db07182fa97c93b1
+  data.tar.gz: 5a27a8983d4c29e9b17b5f5692955a38c26f903c098edb35ad690ed08d88c7c0425daa2bc44aff3742073b380b63eaea41d03dd05a22f2630fca89c4b41a3562

data/Gemfile

@@ -14,3 +14,5 @@ gem "rspec", "~> 3.0"
 
 gem "standard", "~> 1.3"
 gem 'pry', '~> 0.14.2'
+
+gem "yard-doctest", "~> 0.1.17"

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    tiktoken_ruby (0.0.2)
+    tiktoken_ruby (0.0.3)
 
 GEM
   remote: https://rubygems.org/
@@ -12,6 +12,7 @@ GEM
     json (2.6.3)
     language_server-protocol (3.17.0.3)
     method_source (1.0.0)
+    minitest (5.18.0)
     parallel (1.22.1)
     parser (3.2.1.1)
       ast (~> 2.4.1)
@@ -59,6 +60,12 @@ GEM
       rubocop (= 1.48.1)
       rubocop-performance (= 1.16.0)
     unicode-display_width (2.4.2)
+    webrick (1.7.0)
+    yard (0.9.28)
+      webrick (~> 1.7.0)
+    yard-doctest (0.1.17)
+      minitest
+      yard
 
 PLATFORMS
   arm64-darwin-22
@@ -72,6 +79,7 @@ DEPENDENCIES
   rspec (~> 3.0)
   standard (~> 1.3)
   tiktoken_ruby!
+  yard-doctest (~> 0.1.17)
 
 BUNDLED WITH
    2.4.6

data/README.md

@@ -15,11 +15,22 @@ If bundler is not being used to manage dependencies, install the gem by executin
     $ gem install tiktoken_ruby
 
 ## Usage
+Usage should be very similar to the python library. here's a simple example
 
+Encode and decode text
 ```ruby
-encoding = Tiktoken::Encoding.r50k_base
-tokens = encoding.encode("Hello world!")
-puts encoding.decode(tokens)
+require 'tiktoken_ruby'
+
+enc = Tiktoken.get_encoding("cl100k_base")
+enc.decode(enc.encode("hello world")) #=> "hello world"
+```
+
+Encoders can also be retrieved by model name
+```ruby
+require 'tiktoken_ruby'
+
+enc = Tiktoken.encoding_for_model("gpt-4")
+enc.encode("hello world").length #=> 2
 ```
 
 ## Development

data/doctest_helper.rb

@@ -0,0 +1 @@
+require 'lib/tiktoken_ruby.rb'

data/lib/tiktoken_ruby/encoding.rb

@@ -1,23 +1,51 @@
 # frozen_string_literal: true
 
-class Tiktoken::Encoding 
-    def self.method_missing(method)
-        Tiktoken::Encoding.new(Tiktoken::BpeFactory.send(method))
+class Tiktoken::Encoding
+    attr_reader :name
+
+    # This returns a new Tiktoken::Encoding instance for the requested encoding
+    # @param encoding [Symbol] The name of the encoding to load
+    # @return [Tiktoken::Encoding] The encoding instance
+    def self.for_name(encoding)
+        Tiktoken::Encoding.new(Tiktoken::BpeFactory.send(encoding.to_sym), encoding.to_sym)
     end
 
-    def initialize(ext_base_bpe)
-        @ext_base_bpe = ext_base_bpe
+    # This returns a Tiktoken::Encoding instance for the requested encoding
+    # It will reuse an existing encoding if it's already been loaded
+    # @param encoding [Symbol] The name of the encoding to load
+    # @return [Tiktoken::Encoding] The encoding instance
+    def self.for_name_cached(encoding)
+        @encodings ||= {}
+        @encodings[encoding.to_sym] ||= Tiktoken::Encoding.for_name(encoding)
     end
 
+    # Encodes the text as a list of integer tokens. This encoding will encode special non text tokens
+    # basically it's unescaped
+    # @param text [String] The text to encode
+    # @return [Array<Integer>] The encoded tokens
     def encode_ordinary(text)
         @ext_base_bpe.encode_ordinary(text)
     end
 
+    # Encodes the text as a list of integer tokens. This encoding will treat special non text tokens
+    # as text unless they're in the allowed_special array. It's basically like the text was escaped
+    # @param text [String] The text to encode
+    # @param allowed_special [Array<String>] An array of special tokens to allow
+    # @return [Array<Integer>] The encoded tokens
     def encode(text, allowed_special: [])
         @ext_base_bpe.encode(text, allowed_special)
     end
 
+    # Decodes the tokens back into text
+    # @param tokens [Array<Integer>] The tokens to decode
+    # @return [String] The decoded text
     def decode(tokens)
         @ext_base_bpe.decode(tokens)
     end
+
+    private
+    def initialize(ext_base_bpe, name)
+        @ext_base_bpe = ext_base_bpe
+        @name = name
+    end
 end

data/lib/tiktoken_ruby/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Tiktoken
-  VERSION = "0.0.2"
+  VERSION = "0.0.3"
 end

data/lib/tiktoken_ruby.rb

@@ -11,5 +11,105 @@ rescue LoadError
 end
 
 module Tiktoken
-  class Error < StandardError; end
+  class << self
+
+    # Returns an encoding by name. If the encoding is not already loaded it will be loaded, but otherwise
+    # it will reuse the instance of that type that was previous loaded
+    # @param name [Symbol|String] The name of the encoding to load
+    # @return [Tiktoken::Encoding] The encoding instance
+    # @example Encode and decode text
+    #   enc = Tiktoken.get_encoding("cl100k_base")
+    #   enc.decode(enc.encode("hello world")) #=> "hello world"
+    def get_encoding(name)
+      name = name.to_sym
+      return nil unless SUPPORTED_ENCODINGS.include?(name)
+
+      Tiktoken::Encoding.for_name_cached(name)
+    end
+
+    # Gets the encoding for an OpenAI model
+    # @param model_name [Symbol|String] The name of the model to get the encoding for
+    # @return [Tiktoken::Encoding] The encoding instance
+    # @example Count tokens for text
+    #   enc = Tiktoken.encoding_for_model("gpt-4")
+    #   enc.encode("hello world").length #=> 2
+    def encoding_for_model(model_name)
+      for prefix in PREFIX_MODELS
+        if model_name.to_s.start_with?("#{prefix}-")
+          model_name = prefix
+          break
+        end
+      end
+
+      encoding_name = MODEL_TO_ENCODING_NAME[model_name.to_sym]
+      return nil unless encoding_name
+
+      get_encoding(encoding_name)
+    end
+
+    # Lists all the encodings that are supported
+    # @return [Array<Symbol>] The list of supported encodings
+    def list_encoding_names
+      SUPPORTED_ENCODINGS
+    end
+
+    # Lists all the models that are supported
+    # @return [Array<Symbol>] The list of supported models
+    def list_model_names
+      MODEL_TO_ENCODING_NAME.keys
+    end
+
+    private
+
+    SUPPORTED_ENCODINGS = [
+      :r50k_base,
+      :p50k_base,
+      :p50k_edit,
+      :cl100k_base,
+    ]
+
+    # taken from the python library here https://github.com/openai/tiktoken/blob/main/tiktoken/model.py#L13-L53
+    # that is also MIT licensed but by OpenAI
+    MODEL_TO_ENCODING_NAME = {
+      "gpt-4": "cl100k_base",
+      "gpt-3.5-turbo": "cl100k_base",
+      # text
+      "text-davinci-003": "p50k_base",
+      "text-davinci-002": "p50k_base",
+      "text-davinci-001": "r50k_base",
+      "text-curie-001": "r50k_base",
+      "text-babbage-001": "r50k_base",
+      "text-ada-001": "r50k_base",
+      "davinci": "r50k_base",
+      "curie": "r50k_base",
+      "babbage": "r50k_base",
+      "ada": "r50k_base",
+      # code
+      "code-davinci-002": "p50k_base",
+      "code-davinci-001": "p50k_base",
+      "code-cushman-002": "p50k_base",
+      "code-cushman-001": "p50k_base",
+      "davinci-codex": "p50k_base",
+      "cushman-codex": "p50k_base",
+      # edit
+      "text-davinci-edit-001": "p50k_edit",
+      "code-davinci-edit-001": "p50k_edit",
+      # embeddings
+      "text-embedding-ada-002": "cl100k_base",
+      # old embeddings
+      "text-similarity-davinci-001": "r50k_base",
+      "text-similarity-curie-001": "r50k_base",
+      "text-similarity-babbage-001": "r50k_base",
+      "text-similarity-ada-001": "r50k_base",
+      "text-search-davinci-doc-001": "r50k_base",
+      "text-search-curie-doc-001": "r50k_base",
+      "text-search-babbage-doc-001": "r50k_base",
+      "text-search-ada-doc-001": "r50k_base",
+      "code-search-babbage-code-001": "r50k_base",
+      "code-search-ada-code-001": "r50k_base",
+    }
+
+    # these are models that have a versioned models that are otherwise identical
+    PREFIX_MODELS = ["gpt-4", "gpt-3.5-turbo"]
+  end
 end

metadata

@@ -1,16 +1,18 @@
 --- !ruby/object:Gem::Specification
 name: tiktoken_ruby
 version: !ruby/object:Gem::Version
-  version: 0.0.2
+  version: 0.0.3
 platform: x86_64-darwin
 authors:
 - IAPark
 autorequire: 
 bindir: exe
 cert_chain: []
-date: 2023-03-19 00:00:00.000000000 Z
+date: 2023-03-21 00:00:00.000000000 Z
 dependencies: []
-description: Unofficial Ruby wrapper for Tiktoken by way of the unofficial rust bindings
+description: An unofficial Ruby wrapper for Tiktoken, a BPE tokenizer written by and
+  used by OpenAI. It can be used to count the number of tokens in text before sending
+  it to OpenAI APIs.
 email:
 - isaac.a.park@gmail.com
 executables: []
@@ -24,6 +26,7 @@ files:
 - LICENSE.txt
 - README.md
 - Rakefile
+- doctest_helper.rb
 - lib/tiktoken_ruby.rb
 - lib/tiktoken_ruby/2.7/tiktoken_ruby.bundle
 - lib/tiktoken_ruby/3.0/tiktoken_ruby.bundle
@@ -38,6 +41,7 @@ licenses:
 metadata:
   homepage_uri: https://github.com/IAPark/tiktoken_ruby
   source_code_uri: https://github.com/IAPark/tiktoken_ruby
+  documentation_uri: https://rubydoc.info/github/IAPark/tiktoken_ruby/main
 post_install_message: 
 rdoc_options: []
 require_paths: