llm.rb 0.10.1 → 0.12.0

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: llm.rb
 version: !ruby/object:Gem::Version
-  version: 0.10.1
+  version: 0.12.0
 platform: ruby
 authors:
 - Antar Azri
@@ -114,14 +114,14 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '1.40'
+        version: '1.50'
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: '1.40'
+        version: '1.50'
 - !ruby/object:Gem::Dependency
   name: vcr
   requirement: !ruby/object:Gem::Requirement
@@ -151,9 +151,9 @@ dependencies:
       - !ruby/object:Gem::Version
         version: '2.8'
 description: llm.rb is a zero-dependency Ruby toolkit for Large Language Models that
-  includes OpenAI, Gemini, Anthropic, DeepSeek, Ollama, and LlamaCpp. The toolkit
-  includes full support for chat, streaming, tool calling, audio, images, files, and
-  JSON Schema generation.
+  includes OpenAI, Gemini, Anthropic, xAI (grok), DeepSeek, Ollama, and LlamaCpp.
+  The toolkit includes full support for chat, streaming, tool calling, audio, images,
+  files, and JSON Schema generation.
 email:
 - azantar@proton.me
 - 0x1eef@proton.me
@@ -171,7 +171,7 @@ files:
 - lib/llm/bot/prompt/respond.rb
 - lib/llm/buffer.rb
 - lib/llm/error.rb
-- lib/llm/event_handler.rb
+- lib/llm/eventhandler.rb
 - lib/llm/eventstream.rb
 - lib/llm/eventstream/event.rb
 - lib/llm/eventstream/parser.rb
@@ -189,7 +189,6 @@ files:
 - lib/llm/json/schema/version.rb
 - lib/llm/message.rb
 - lib/llm/mime.rb
-- lib/llm/model.rb
 - lib/llm/multipart.rb
 - lib/llm/object.rb
 - lib/llm/object/builder.rb
@@ -200,8 +199,7 @@ files:
 - lib/llm/providers/anthropic/format.rb
 - lib/llm/providers/anthropic/format/completion_format.rb
 - lib/llm/providers/anthropic/models.rb
-- lib/llm/providers/anthropic/response_parser.rb
-- lib/llm/providers/anthropic/response_parser/completion_parser.rb
+- lib/llm/providers/anthropic/response/completion.rb
 - lib/llm/providers/anthropic/stream_parser.rb
 - lib/llm/providers/deepseek.rb
 - lib/llm/providers/deepseek/format.rb
@@ -214,8 +212,10 @@ files:
 - lib/llm/providers/gemini/format/completion_format.rb
 - lib/llm/providers/gemini/images.rb
 - lib/llm/providers/gemini/models.rb
-- lib/llm/providers/gemini/response_parser.rb
-- lib/llm/providers/gemini/response_parser/completion_parser.rb
+- lib/llm/providers/gemini/response/completion.rb
+- lib/llm/providers/gemini/response/embedding.rb
+- lib/llm/providers/gemini/response/file.rb
+- lib/llm/providers/gemini/response/image.rb
 - lib/llm/providers/gemini/stream_parser.rb
 - lib/llm/providers/llamacpp.rb
 - lib/llm/providers/ollama.rb
@@ -223,8 +223,8 @@ files:
 - lib/llm/providers/ollama/format.rb
 - lib/llm/providers/ollama/format/completion_format.rb
 - lib/llm/providers/ollama/models.rb
-- lib/llm/providers/ollama/response_parser.rb
-- lib/llm/providers/ollama/response_parser/completion_parser.rb
+- lib/llm/providers/ollama/response/completion.rb
+- lib/llm/providers/ollama/response/embedding.rb
 - lib/llm/providers/ollama/stream_parser.rb
 - lib/llm/providers/openai.rb
 - lib/llm/providers/openai/audio.rb
@@ -237,34 +237,25 @@ files:
 - lib/llm/providers/openai/images.rb
 - lib/llm/providers/openai/models.rb
 - lib/llm/providers/openai/moderations.rb
-- lib/llm/providers/openai/response_parser.rb
-- lib/llm/providers/openai/response_parser/completion_parser.rb
-- lib/llm/providers/openai/response_parser/respond_parser.rb
+- lib/llm/providers/openai/response/audio.rb
+- lib/llm/providers/openai/response/completion.rb
+- lib/llm/providers/openai/response/embedding.rb
+- lib/llm/providers/openai/response/file.rb
+- lib/llm/providers/openai/response/image.rb
+- lib/llm/providers/openai/response/moderations.rb
+- lib/llm/providers/openai/response/responds.rb
 - lib/llm/providers/openai/responses.rb
 - lib/llm/providers/openai/stream_parser.rb
-- lib/llm/providers/voyageai.rb
-- lib/llm/providers/voyageai/error_handler.rb
-- lib/llm/providers/voyageai/response_parser.rb
+- lib/llm/providers/openai/vector_stores.rb
+- lib/llm/providers/xai.rb
+- lib/llm/providers/xai/images.rb
 - lib/llm/response.rb
-- lib/llm/response/audio.rb
-- lib/llm/response/audio_transcription.rb
-- lib/llm/response/audio_translation.rb
-- lib/llm/response/completion.rb
-- lib/llm/response/download_file.rb
-- lib/llm/response/embedding.rb
-- lib/llm/response/file.rb
-- lib/llm/response/filelist.rb
-- lib/llm/response/image.rb
-- lib/llm/response/modellist.rb
-- lib/llm/response/moderationlist.rb
-- lib/llm/response/moderationlist/moderation.rb
-- lib/llm/response/respond.rb
 - lib/llm/utils.rb
 - lib/llm/version.rb
 - llm.gemspec
 homepage: https://github.com/llmrb/llm
 licenses:
-- 0BSDL
+- 0BSD
 metadata:
   homepage_uri: https://github.com/llmrb/llm
   source_code_uri: https://github.com/llmrb/llm
@@ -282,10 +273,10 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.7.1
+rubygems_version: 3.6.9
 specification_version: 4
 summary: llm.rb is a zero-dependency Ruby toolkit for Large Language Models that includes
-  OpenAI, Gemini, Anthropic, DeepSeek, Ollama, and LlamaCpp. The toolkit includes
-  full support for chat, streaming, tool calling, audio, images, files, and JSON Schema
-  generation.
+  OpenAI, Gemini, Anthropic, xAI (grok), DeepSeek, Ollama, and LlamaCpp. The toolkit
+  includes full support for chat, streaming, tool calling, audio, images, files, and
+  JSON Schema generation.
 test_files: []

data/lib/llm/model.rb

@@ -1,32 +0,0 @@
-# frozen_string_literal: true
-
-##
-# The {LLM::Model LLM::Model} class represents an LLM model that
-# is available to use. Its properties are delegated to the underlying
-# response body, and vary by provider.
-class LLM::Model < LLM::Object
-  ##
-  # Returns a subclass of {LLM::Provider LLM::Provider}
-  # @return [LLM::Provider]
-  attr_accessor :provider
-
-  ##
-  # Returns the model ID
-  # @return [String]
-  def id
-    case @provider.class.to_s
-    when "LLM::Ollama"
-      self["name"]
-    when "LLM::Gemini"
-      self["name"].sub(%r|\Amodels/|, "")
-    else
-      self["id"]
-    end
-  end
-
-  ##
-  # @return [String]
-  def to_json(*)
-    id.to_json(*)
-  end
-end

data/lib/llm/providers/anthropic/response_parser/completion_parser.rb

@@ -1,51 +0,0 @@
-# frozen_string_literal: true
-
-module LLM::Anthropic::ResponseParser
-  ##
-  # @private
-  class CompletionParser
-    def initialize(body)
-      @body = LLM::Object.from_hash(body)
-    end
-
-    def format(response)
-      {
-        model:,
-        prompt_tokens:,
-        completion_tokens:,
-        total_tokens:,
-        choices: format_choices(response)
-      }
-    end
-
-    private
-
-    def format_choices(response)
-      texts.map.with_index do |choice, index|
-        extra = {index:, response:, tool_calls: format_tool_calls(tools), original_tool_calls: tools}
-        LLM::Message.new(role, choice.text, extra)
-      end
-    end
-
-    def format_tool_calls(tools)
-      (tools || []).filter_map do |tool|
-        tool = {
-          id: tool.id,
-          name: tool.name,
-          arguments: tool.input
-        }
-        LLM::Object.new(tool)
-      end
-    end
-
-    def body = @body
-    def role = body.role
-    def model = body.model
-    def prompt_tokens = body.usage&.input_tokens
-    def completion_tokens = body.usage&.output_tokens
-    def total_tokens = body.usage&.total_tokens
-    def parts = body.content
-    def texts = parts.select { _1["type"] == "text" }
-    def tools = parts.select { _1["type"] == "tool_use" }
-  end
-end

data/lib/llm/providers/anthropic/response_parser.rb

@@ -1,24 +0,0 @@
-# frozen_string_literal: true
-
-class LLM::Anthropic
-  ##
-  # @private
-  module ResponseParser
-    require_relative "response_parser/completion_parser"
-    def parse_embedding(body)
-      {
-        model: body["model"],
-        embeddings: body["data"].map { _1["embedding"] },
-        total_tokens: body.dig("usage", "total_tokens")
-      }
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_completion(body)
-      CompletionParser.new(body).format(self)
-    end
-  end
-end

data/lib/llm/providers/gemini/response_parser.rb

@@ -1,46 +0,0 @@
-# frozen_string_literal: true
-
-class LLM::Gemini
-  ##
-  # @private
-  module ResponseParser
-    require_relative "response_parser/completion_parser"
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_completion(body)
-      CompletionParser.new(body).format(self)
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_embedding(body)
-      {
-        model: "text-embedding-004",
-        embeddings: body.dig("embedding", "values")
-      }
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_image(body)
-      {
-        urls: [],
-        images: body["candidates"].flat_map do |c|
-          parts = c["content"]["parts"]
-          parts.filter_map do
-            data = _1.dig("inlineData", "data")
-            next unless data
-            StringIO.new(data.unpack1("m0"))
-          end
-        end
-      }
-    end
-  end
-end

data/lib/llm/providers/ollama/response_parser/completion_parser.rb

@@ -1,42 +0,0 @@
-# frozen_string_literal: true
-
-module LLM::Ollama::ResponseParser
-  ##
-  # @private
-  class CompletionParser
-    def initialize(body)
-      @body = LLM::Object.from_hash(body)
-    end
-
-    def format(response)
-      {
-        model:,
-        choices: [format_choices(response)],
-        prompt_tokens:,
-        completion_tokens:
-      }
-    end
-
-    private
-
-    def format_choices(response)
-      role, content, calls = message.to_h.values_at(:role, :content, :tool_calls)
-      extra = {response:, tool_calls: format_tool_calls(calls)}
-      LLM::Message.new(role, content, extra)
-    end
-
-    def format_tool_calls(tools)
-      return [] unless tools
-      tools.filter_map do |tool|
-        next unless tool["function"]
-        LLM::Object.new(tool["function"])
-      end
-    end
-
-    def body = @body
-    def model = body.model
-    def prompt_tokens = body.prompt_eval_count
-    def completion_tokens = body.eval_count
-    def message = body.message
-  end
-end

data/lib/llm/providers/ollama/response_parser.rb

@@ -1,30 +0,0 @@
-# frozen_string_literal: true
-
-class LLM::Ollama
-  ##
-  # @private
-  module ResponseParser
-    require_relative "response_parser/completion_parser"
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_completion(body)
-      CompletionParser.new(body).format(self)
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_embedding(body)
-      {
-        model: body["model"],
-        embeddings: body["data"].map { _1["embedding"] },
-        prompt_tokens: body.dig("usage", "prompt_tokens"),
-        total_tokens: body.dig("usage", "total_tokens")
-      }
-    end
-  end
-end

data/lib/llm/providers/openai/response_parser.rb

@@ -1,65 +0,0 @@
-# frozen_string_literal: true
-
-class LLM::OpenAI
-  ##
-  # @private
-  module ResponseParser
-    require_relative "response_parser/completion_parser"
-    require_relative "response_parser/respond_parser"
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_completion(body)
-      CompletionParser.new(body).format(self)
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_respond_response(body)
-      RespondParser.new(body).format(self)
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_moderation_list(body)
-      {
-        id: body["id"],
-        model: body["model"],
-        moderations: body["results"].map { LLM::Response::ModerationList::Moderation.new(_1) }
-      }
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_embedding(body)
-      {
-        model: body["model"],
-        embeddings: body["data"].map { _1["embedding"] },
-        prompt_tokens: body.dig("usage", "prompt_tokens"),
-        total_tokens: body.dig("usage", "total_tokens")
-      }
-    end
-
-    ##
-    # @param [Hash] body
-    #  The response body from the LLM provider
-    # @return [Hash]
-    def parse_image(body)
-      {
-        urls: body["data"].filter_map { _1["url"] },
-        images: body["data"].filter_map do
-          next unless _1["b64_json"]
-          StringIO.new(_1["b64_json"].unpack1("m0"))
-        end
-      }
-    end
-  end
-end

data/lib/llm/providers/voyageai/error_handler.rb

@@ -1,32 +0,0 @@
-# frozen_string_literal: true
-
-class LLM::VoyageAI
-  class ErrorHandler
-    ##
-    # @return [Net::HTTPResponse]
-    #  Non-2XX response from the server
-    attr_reader :res
-
-    ##
-    # @param [Net::HTTPResponse] res
-    #  The response from the server
-    # @return [LLM::OpenAI::ErrorHandler]
-    def initialize(res)
-      @res = res
-    end
-
-    ##
-    # @raise [LLM::Error]
-    #  Raises a subclass of {LLM::Error LLM::Error}
-    def raise_error!
-      case res
-      when Net::HTTPUnauthorized
-        raise LLM::UnauthorizedError.new { _1.response = res }, "Authentication error"
-      when Net::HTTPTooManyRequests
-        raise LLM::RateLimitError.new { _1.response = res }, "Too many requests"
-      else
-        raise LLM::ResponseError.new { _1.response = res }, "Unexpected response"
-      end
-    end
-  end
-end

data/lib/llm/providers/voyageai/response_parser.rb

@@ -1,13 +0,0 @@
-# frozen_string_literal: true
-
-class LLM::VoyageAI
-  module ResponseParser
-    def parse_embedding(body)
-      {
-        model: body["model"],
-        embeddings: body["data"].map { _1["embedding"] },
-        total_tokens: body.dig("usage", "total_tokens")
-      }
-    end
-  end
-end

data/lib/llm/providers/voyageai.rb

@@ -1,44 +0,0 @@
-# frozen_string_literal: true
-
-module LLM
-  class VoyageAI < Provider
-    require_relative "voyageai/error_handler"
-    require_relative "voyageai/response_parser"
-    HOST = "api.voyageai.com"
-
-    ##
-    # @param key (see LLM::Provider#initialize)
-    def initialize(**)
-      super(host: HOST, **)
-    end
-
-    ##
-    # Provides an embedding via VoyageAI per
-    # [Anthropic's recommendation](https://docs.anthropic.com/en/docs/build-with-claude/embeddings)
-    # @param input (see LLM::Provider#embed)
-    # @return (see LLM::Provider#embed)
-    def embed(input, model: "voyage-2", **params)
-      req = Net::HTTP::Post.new("/v1/embeddings", headers)
-      req.body = JSON.dump({input:, model:}.merge!(params))
-      res = execute(request: req)
-      Response::Embedding.new(res).extend(response_parser)
-    end
-
-    private
-
-    def headers
-      {
-        "Content-Type" => "application/json",
-        "Authorization" => "Bearer #{@key}"
-      }
-    end
-
-    def response_parser
-      LLM::VoyageAI::ResponseParser
-    end
-
-    def error_handler
-      LLM::VoyageAI::ErrorHandler
-    end
-  end
-end

data/lib/llm/response/audio.rb

@@ -1,13 +0,0 @@
-# frozen_string_literal: true
-
-module LLM
-  ##
-  # The {LLM::Response::Audio LLM::Response::Audio} class represents an
-  # audio file that has been returned by a provider. It wraps an IO object
-  # that can be used to read the contents of an audio stream (as binary data).
-  class Response::Audio < Response
-    ##
-    # @return [StringIO]
-    attr_accessor :audio
-  end
-end

data/lib/llm/response/audio_transcription.rb

@@ -1,14 +0,0 @@
-# frozen_string_literal: true
-
-module LLM
-  ##
-  # The {LLM::Response::AudioTranscription LLM::Response::AudioTranscription}
-  # class represents an audio transcription that has been returned by
-  # a provider (eg OpenAI, Gemini, etc)
-  class Response::AudioTranscription < Response
-    ##
-    # Returns the text of the transcription
-    # @return [String]
-    attr_accessor :text
-  end
-end

data/lib/llm/response/audio_translation.rb

@@ -1,14 +0,0 @@
-# frozen_string_literal: true
-
-module LLM
-  ##
-  # The {LLM::Response::AudioTranslation LLM::Response::AudioTranslation}
-  # class represents an audio translation that has been returned by
-  # a provider (eg OpenAI, Gemini, etc)
-  class Response::AudioTranslation < Response
-    ##
-    # Returns the text of the translation
-    # @return [String]
-    attr_accessor :text
-  end
-end

data/lib/llm/response/completion.rb

@@ -1,51 +0,0 @@
-# frozen_string_literal: true
-
-module LLM
-  class Response::Completion < Response
-    ##
-    # @return [String]
-    #   Returns the model name used for the completion
-    def model
-      parsed[:model]
-    end
-
-    ##
-    # @return [Array<LLM::Message>]
-    #   Returns an array of messages
-    def choices
-      parsed[:choices]
-    end
-    alias_method :messages, :choices
-
-    ##
-    # @return [Integer]
-    #   Returns the count of prompt tokens
-    def prompt_tokens
-      parsed[:prompt_tokens]
-    end
-
-    ##
-    # @return [Integer]
-    #   Returns the count of completion tokens
-    def completion_tokens
-      parsed[:completion_tokens]
-    end
-
-    ##
-    # @return [Integer]
-    #   Returns the total count of tokens
-    def total_tokens
-      prompt_tokens + completion_tokens
-    end
-
-    private
-
-    ##
-    # @private
-    # @return [Hash]
-    #   Returns the parsed completion response from the provider
-    def parsed
-      @parsed ||= parse_completion(body)
-    end
-  end
-end

data/lib/llm/response/download_file.rb

@@ -1,15 +0,0 @@
-# frozen_string_literal: true
-
-module LLM
-  ##
-  # The {LLM::Response::DownloadFile LLM::Response::DownloadFile} class
-  # represents the contents of a file that has been returned by a
-  # provider. It wraps an IO object that can be used to read the file
-  # contents.
-  class Response::DownloadFile < Response
-    ##
-    # Returns a StringIO object
-    # @return [StringIO]
-    attr_accessor :file
-  end
-end

data/lib/llm/response/embedding.rb

@@ -1,23 +0,0 @@
-# frozen_string_literal: true
-
-module LLM
-  class Response::Embedding < Response
-    def model
-      parsed[:model]
-    end
-
-    def embeddings
-      parsed[:embeddings]
-    end
-
-    def total_tokens
-      parsed[:total_tokens]
-    end
-
-    private
-
-    def parsed
-      @parsed ||= parse_embedding(body)
-    end
-  end
-end