ruby_llm 0.1.0.pre30 → 0.1.0.pre33

data/lib/ruby_llm/model_capabilities/anthropic.rb

@@ -1,79 +0,0 @@
-# frozen_string_literal: true
-
-module RubyLLM
-  module ModelCapabilities
-    # Determines capabilities and pricing for Anthropic models
-    module Anthropic
-      module_function
-
-      def determine_context_window(model_id)
-        case model_id
-        when /claude-3/ then 200_000
-        else 100_000
-        end
-      end
-
-      def determine_max_tokens(model_id)
-        case model_id
-        when /claude-3-5/ then 8_192
-        else 4_096
-        end
-      end
-
-      def get_input_price(model_id)
-        PRICES.dig(model_family(model_id), :input) || default_input_price
-      end
-
-      def get_output_price(model_id)
-        PRICES.dig(model_family(model_id), :output) || default_output_price
-      end
-
-      def supports_vision?(model_id)
-        return false if model_id.match?(/claude-3-5-haiku/)
-        return false if model_id.match?(/claude-[12]/)
-
-        true
-      end
-
-      def supports_functions?(model_id)
-        model_id.include?('claude-3')
-      end
-
-      def supports_json_mode?(model_id)
-        model_id.include?('claude-3')
-      end
-
-      def model_family(model_id)
-        case model_id
-        when /claude-3-5-sonnet/  then :claude35_sonnet
-        when /claude-3-5-haiku/   then :claude35_haiku
-        when /claude-3-opus/      then :claude3_opus
-        when /claude-3-sonnet/    then :claude3_sonnet
-        when /claude-3-haiku/     then :claude3_haiku
-        else :claude2
-        end
-      end
-
-      def model_type(_)
-        'chat'
-      end
-
-      PRICES = {
-        claude35_sonnet: { input: 3.0, output: 15.0 }, # $3.00/$15.00 per million tokens
-        claude35_haiku: { input: 0.80, output: 4.0 }, # $0.80/$4.00 per million tokens
-        claude3_opus: { input: 15.0, output: 75.0 }, # $15.00/$75.00 per million tokens
-        claude3_sonnet: { input: 3.0, output: 15.0 }, # $3.00/$15.00 per million tokens
-        claude3_haiku: { input: 0.25, output: 1.25 }, # $0.25/$1.25 per million tokens
-        claude2: { input: 3.0,  output: 15.0 } # Default pricing for Claude 2.x models
-      }.freeze
-
-      def default_input_price
-        3.0
-      end
-
-      def default_output_price
-        15.0
-      end
-    end
-  end
-end

data/lib/ruby_llm/model_capabilities/deepseek.rb

@@ -1,132 +0,0 @@
-# frozen_string_literal: true
-
-module RubyLLM
-  module ModelCapabilities
-    # Determines capabilities and pricing for DeepSeek models
-    module DeepSeek
-      module_function
-
-      # Returns the context window size for the given model
-      # @param model_id [String] the model identifier
-      # @return [Integer] the context window size in tokens
-      def context_window_for(model_id)
-        case model_id
-        when /deepseek-(?:chat|reasoner)/ then 64_000
-        else 32_768 # Sensible default
-        end
-      end
-
-      # Returns the maximum output tokens for the given model
-      # @param model_id [String] the model identifier
-      # @return [Integer] the maximum output tokens
-      def max_tokens_for(model_id)
-        case model_id
-        when /deepseek-(?:chat|reasoner)/ then 8_192
-        else 4_096 # Default if max_tokens not specified
-        end
-      end
-
-      # Returns the input price per million tokens for the given model
-      # @param model_id [String] the model identifier
-      # @return [Float] the price per million tokens
-      def input_price_for(model_id)
-        PRICES.dig(model_family(model_id), :input_miss) || default_input_price
-      end
-
-      # Returns the output price per million tokens for the given model
-      # @param model_id [String] the model identifier
-      # @return [Float] the price per million tokens
-      def output_price_for(model_id)
-        PRICES.dig(model_family(model_id), :output) || default_output_price
-      end
-
-      # Returns the cache hit price per million tokens for the given model
-      # @param model_id [String] the model identifier
-      # @return [Float] the price per million tokens
-      def cache_hit_price_for(model_id)
-        PRICES.dig(model_family(model_id), :input_hit) || default_cache_hit_price
-      end
-
-      # Determines if the model supports vision capabilities
-      # @param model_id [String] the model identifier
-      # @return [Boolean] true if the model supports vision
-      def supports_vision?(_model_id)
-        false # DeepSeek models don't currently support vision
-      end
-
-      # Determines if the model supports function calling
-      # @param model_id [String] the model identifier
-      # @return [Boolean] true if the model supports function calling
-      def supports_functions?(model_id)
-        model_id.match?(/deepseek-chat/) # Only deepseek-chat supports function calling
-      end
-
-      # Determines if the model supports JSON mode
-      # @param model_id [String] the model identifier
-      # @return [Boolean] true if the model supports JSON mode
-      def supports_json_mode?(model_id)
-        model_id.match?(/deepseek-chat/) # Only deepseek-chat supports JSON mode
-      end
-
-      # Formats the model ID into a display name
-      # @param model_id [String] the model identifier
-      # @return [String] the formatted display name
-      def format_display_name(model_id)
-        case model_id
-        when 'deepseek-chat' then 'DeepSeek V3'
-        when 'deepseek-reasoner' then 'DeepSeek R1'
-        else
-          model_id.split('-')
-                  .map(&:capitalize)
-                  .join(' ')
-        end
-      end
-
-      # Returns the model type
-      # @param model_id [String] the model identifier
-      # @return [String] the model type
-      def model_type(_model_id)
-        'chat' # All DeepSeek models are chat models
-      end
-
-      # Returns the model family for pricing purposes
-      # @param model_id [String] the model identifier
-      # @return [String] the model family identifier
-      def model_family(model_id)
-        case model_id
-        when /deepseek-chat/ then :chat
-        when /deepseek-reasoner/ then :reasoner
-        else :chat # Default to chat family
-        end
-      end
-
-      # Pricing information for DeepSeek models (USD per 1M tokens)
-      PRICES = {
-        chat: {
-          input_hit: 0.07,   # $0.07 per million tokens on cache hit
-          input_miss: 0.27,  # $0.27 per million tokens on cache miss
-          output: 1.10       # $1.10 per million tokens output
-        },
-        reasoner: {
-          input_hit: 0.14,   # $0.14 per million tokens on cache hit
-          input_miss: 0.55,  # $0.55 per million tokens on cache miss
-          output: 2.19       # $2.19 per million tokens output
-        }
-      }.freeze
-
-      private
-
-      def default_input_price
-        0.27 # Default to chat cache miss price
-      end
-
-      def default_output_price
-        1.10 # Default to chat output price
-      end
-
-      def default_cache_hit_price
-        0.07 # Default to chat cache hit price
-      end
-    end
-  end
-end

data/lib/ruby_llm/model_capabilities/gemini.rb

@@ -1,190 +0,0 @@
-# frozen_string_literal: true
-
-module RubyLLM
-  module ModelCapabilities
-    # Determines capabilities and pricing for Google Gemini models
-    module Gemini # rubocop:disable Metrics/ModuleLength
-      extend self
-
-      def context_window_for(model_id)
-        case model_id
-        when /gemini-2\.0-flash/, /gemini-1\.5-flash/ then 1_048_576
-        when /gemini-1\.5-pro/ then 2_097_152
-        when /text-embedding/, /embedding-001/ then 2_048
-        when /aqa/ then 7_168
-        else 32_768 # Sensible default for unknown models
-        end
-      end
-
-      def max_tokens_for(model_id)
-        case model_id
-        when /gemini-2\.0-flash/, /gemini-1\.5/ then 8_192
-        when /text-embedding/, /embedding-001/ then 768 # Output dimension size for embeddings
-        when /aqa/ then 1_024
-        else 4_096 # Sensible default
-        end
-      end
-
-      def input_price_for(model_id)
-        base_price = PRICES.dig(pricing_family(model_id), :input) || default_input_price
-        return base_price unless long_context_model?(model_id)
-
-        # Double the price for prompts longer than 128k tokens
-        context_length(model_id) > 128_000 ? base_price * 2 : base_price
-      end
-
-      def output_price_for(model_id)
-        base_price = PRICES.dig(pricing_family(model_id), :output) || default_output_price
-        return base_price unless long_context_model?(model_id)
-
-        # Double the price for prompts longer than 128k tokens
-        context_length(model_id) > 128_000 ? base_price * 2 : base_price
-      end
-
-      def supports_vision?(model_id)
-        return false if model_id.match?(/text-embedding|embedding-001|aqa/)
-        return false if model_id.match?(/gemini-1\.0/)
-
-        model_id.match?(/gemini-[12]\.[05]/)
-      end
-
-      def supports_functions?(model_id)
-        return false if model_id.match?(/text-embedding|embedding-001|aqa/)
-        return false if model_id.match?(/flash-lite/)
-        return false if model_id.match?(/gemini-1\.0/)
-
-        model_id.match?(/gemini-[12]\.[05]-(?:pro|flash)(?!-lite)/)
-      end
-
-      def supports_json_mode?(model_id)
-        return false if model_id.match?(/text-embedding|embedding-001|aqa/)
-        return false if model_id.match?(/flash-lite/)
-        return false if model_id.match?(/gemini-1\.0/)
-
-        model_id.match?(/gemini-1\.5/)
-      end
-
-      def format_display_name(model_id)
-        model_id
-          .delete_prefix('models/')
-          .split('-')
-          .map(&:capitalize)
-          .join(' ')
-          .gsub(/(\d+\.\d+)/, ' \1') # Add space before version numbers
-          .gsub(/\s+/, ' ')          # Clean up multiple spaces
-          .gsub(/Aqa/, 'AQA')        # Special case for AQA
-          .strip
-      end
-
-      def supports_caching?(model_id)
-        return false if model_id.match?(/flash-lite|gemini-1\.0/)
-
-        model_id.match?(/gemini-[12]\.[05]/)
-      end
-
-      def supports_tuning?(model_id)
-        model_id.match?(/gemini-1\.5-flash/)
-      end
-
-      def supports_audio?(model_id)
-        model_id.match?(/gemini-[12]\.[05]/)
-      end
-
-      def model_type(model_id)
-        case model_id
-        when /text-embedding|embedding/ then 'embedding'
-        when /imagen/ then 'image'
-        else 'chat'
-        end
-      end
-
-      def model_family(model_id) # rubocop:disable Metrics/CyclomaticComplexity,Metrics/MethodLength
-        case model_id
-        when /gemini-2\.0-flash-lite/ then 'gemini20_flash_lite'
-        when /gemini-2\.0-flash/ then 'gemini20_flash'
-        when /gemini-1\.5-flash-8b/ then 'gemini15_flash_8b'
-        when /gemini-1\.5-flash/ then 'gemini15_flash'
-        when /gemini-1\.5-pro/ then 'gemini15_pro'
-        when /gemini-1\.0-pro/ then 'gemini10_pro'
-        when /text-embedding-004/ then 'embedding4'
-        when /embedding-001/ then 'embedding1'
-        when /aqa/ then 'aqa'
-        else 'other'
-        end
-      end
-
-      def pricing_family(model_id) # rubocop:disable Metrics/CyclomaticComplexity
-        case model_id
-        when /gemini-2\.0-flash-lite/ then :flash_lite_2 # rubocop:disable Naming/VariableNumber
-        when /gemini-2\.0-flash/ then :flash_2 # rubocop:disable Naming/VariableNumber
-        when /gemini-1\.5-flash-8b/ then :flash_8b
-        when /gemini-1\.5-flash/ then :flash
-        when /gemini-1\.5-pro/ then :pro
-        when /gemini-1\.0-pro/ then :pro_1_0 # rubocop:disable Naming/VariableNumber
-        when /text-embedding|embedding/ then :embedding
-        else :base
-        end
-      end
-
-      private
-
-      def long_context_model?(model_id)
-        model_id.match?(/gemini-1\.5-(?:pro|flash)/)
-      end
-
-      def context_length(model_id)
-        context_window_for(model_id)
-      end
-
-      PRICES = {
-        flash_2: { # Gemini 2.0 Flash # rubocop:disable Naming/VariableNumber
-          input: 0.10,
-          output: 0.40,
-          audio_input: 0.70,
-          cache: 0.025,
-          cache_storage: 1.00
-        },
-        flash_lite_2: { # Gemini 2.0 Flash Lite # rubocop:disable Naming/VariableNumber
-          input: 0.075,
-          output: 0.30,
-          cache: 0.01875,
-          cache_storage: 1.00
-        },
-        flash: { # Gemini 1.5 Flash
-          input: 0.075,
-          output: 0.30,
-          cache: 0.01875,
-          cache_storage: 1.00
-        },
-        flash_8b: { # Gemini 1.5 Flash 8B
-          input: 0.0375,
-          output: 0.15,
-          cache: 0.01,
-          cache_storage: 0.25
-        },
-        pro: { # Gemini 1.5 Pro
-          input: 1.25,
-          output: 5.0,
-          cache: 0.3125,
-          cache_storage: 4.50
-        },
-        pro_1_0: { # Gemini 1.0 Pro # rubocop:disable Naming/VariableNumber
-          input: 0.50,
-          output: 1.50
-        },
-        embedding: { # Text Embedding models
-          input: 0.00,
-          output: 0.00
-        }
-      }.freeze
-
-      def default_input_price
-        0.075 # Default to Flash pricing
-      end
-
-      def default_output_price
-        0.30  # Default to Flash pricing
-      end
-    end
-  end
-end

data/lib/ruby_llm/model_capabilities/openai.rb

@@ -1,189 +0,0 @@
-# frozen_string_literal: true
-
-module RubyLLM
-  module ModelCapabilities
-    # Determines capabilities and pricing for OpenAI models
-    module OpenAI # rubocop:disable Metrics/ModuleLength
-      extend self
-
-      def context_window_for(model_id)
-        case model_id
-        when /o[13]-mini/, /o3-mini-2025/         then 200_000
-        when /o1-2024/                            then 200_000
-        when /gpt-4o/, /gpt-4-turbo/             then 128_000
-        when /gpt-4-0[0-9]{3}/                   then 8_192
-        when /gpt-3.5-turbo-instruct/            then 4_096
-        when /gpt-3.5/                           then 16_385
-        else 4_096
-        end
-      end
-
-      def max_tokens_for(model_id) # rubocop:disable Metrics/CyclomaticComplexity
-        case model_id
-        when /o1-2024/, /o3-mini/                then 100_000
-        when /o1-mini-2024/                      then 65_536
-        when /gpt-4o-2024-05-13/                then 4_096
-        when /gpt-4o/, /gpt-4o-mini/            then 16_384
-        when /gpt-4o-realtime/                  then 4_096
-        when /gpt-4-0[0-9]{3}/                  then 8_192
-        when /gpt-3.5-turbo/                    then 4_096
-        else 4_096
-        end
-      end
-
-      def input_price_for(model_id)
-        PRICES.dig(model_family(model_id), :input) || default_input_price
-      end
-
-      def output_price_for(model_id)
-        PRICES.dig(model_family(model_id), :output) || default_output_price
-      end
-
-      def supports_vision?(model_id)
-        model_id.match?(/gpt-4o|o1/) || model_id.match?(/gpt-4-(?!0314|0613)/)
-      end
-
-      def supports_functions?(model_id)
-        !model_id.include?('instruct')
-      end
-
-      def supports_audio?(model_id)
-        model_id.match?(/audio-preview|realtime-preview|whisper|tts/)
-      end
-
-      def supports_json_mode?(model_id)
-        model_id.match?(/gpt-4-\d{4}-preview/) ||
-          model_id.include?('turbo') ||
-          model_id.match?(/gpt-3.5-turbo-(?!0301|0613)/)
-      end
-
-      def format_display_name(model_id)
-        model_id.then { |id| humanize(id) }
-                .then { |name| apply_special_formatting(name) }
-      end
-
-      def model_type(model_id)
-        case model_id
-        when /text-embedding|embedding/ then 'embedding'
-        when /dall-e/ then 'image'
-        when /tts|whisper/ then 'audio'
-        when /omni-moderation/ then 'moderation'
-        else 'chat'
-        end
-      end
-
-      def supports_structured_output?(model_id)
-        model_id.match?(/gpt-4o|o[13]-mini|o1/)
-      end
-
-      def model_family(model_id) # rubocop:disable Metrics/AbcSize,Metrics/CyclomaticComplexity,Metrics/MethodLength
-        case model_id
-        when /o3-mini/ then 'o3_mini'
-        when /o1-mini/ then 'o1_mini'
-        when /o1/ then 'o1'
-        when /gpt-4o-audio/ then 'gpt4o_audio'
-        when /gpt-4o-realtime/ then 'gpt4o_realtime'
-        when /gpt-4o-mini-audio/ then 'gpt4o_mini_audio'
-        when /gpt-4o-mini-realtime/ then 'gpt4o_mini_realtime'
-        when /gpt-4o-mini/ then 'gpt4o_mini'
-        when /gpt-4o/ then 'gpt4o'
-        when /gpt-4-turbo/ then 'gpt4_turbo'
-        when /gpt-4/ then 'gpt4'
-        when /gpt-3.5-turbo-instruct/ then 'gpt35_instruct'
-        when /gpt-3.5/ then 'gpt35'
-        when /dall-e-3/ then 'dalle3'
-        when /dall-e-2/ then 'dalle2'
-        when /text-embedding-3-large/ then 'embedding3_large'
-        when /text-embedding-3-small/ then 'embedding3_small'
-        when /text-embedding-ada/ then 'embedding2'
-        when /tts-1-hd/ then 'tts1_hd'
-        when /tts-1/ then 'tts1'
-        when /whisper/ then 'whisper1'
-        when /omni-moderation/ then 'moderation'
-        when /babbage/ then 'babbage'
-        when /davinci/ then 'davinci'
-        else 'other'
-        end
-      end
-
-      private
-
-      PRICES = {
-        o1: { input: 15.0, cached_input: 7.5, output: 60.0 },
-        o1_mini: { input: 1.10, cached_input: 0.55, output: 4.40 },
-        o3_mini: { input: 1.10, cached_input: 0.55, output: 4.40 },
-        gpt4o: { input: 2.50, cached_input: 1.25, output: 10.0 },
-        gpt4o_audio: {
-          text_input: 2.50,
-          audio_input: 40.0,
-          text_output: 10.0,
-          audio_output: 80.0
-        },
-        gpt4o_realtime: {
-          text_input: 5.0,
-          cached_text_input: 2.50,
-          audio_input: 40.0,
-          cached_audio_input: 2.50,
-          text_output: 20.0,
-          audio_output: 80.0
-        },
-        gpt4o_mini: { input: 0.15, cached_input: 0.075, output: 0.60 },
-        gpt4o_mini_audio: {
-          text_input: 0.15,
-          audio_input: 10.0,
-          text_output: 0.60,
-          audio_output: 20.0
-        },
-        gpt4o_mini_realtime: {
-          text_input: 0.60,
-          cached_text_input: 0.30,
-          audio_input: 10.0,
-          cached_audio_input: 0.30,
-          text_output: 2.40,
-          audio_output: 20.0
-        },
-        gpt4_turbo: { input: 10.0, output: 30.0 },
-        gpt4: { input: 30.0, output: 60.0 },
-        gpt35: { input: 0.50, output: 1.50 },
-        gpt35_instruct: { input: 1.50, output: 2.0 },
-        embedding3_large: { price: 0.13 },
-        embedding3_small: { price: 0.02 },
-        embedding2: { price: 0.10 },
-        davinci: { input: 2.0, output: 2.0 },
-        babbage: { input: 0.40, output: 0.40 },
-        tts1: { price: 15.0 },
-        tts1_hd: { price: 30.0 },
-        whisper1: { price: 0.006 }
-      }.freeze
-
-      def default_input_price
-        0.50
-      end
-
-      def default_output_price
-        1.50
-      end
-
-      def humanize(id)
-        id.tr('-', ' ')
-          .split(' ')
-          .map(&:capitalize)
-          .join(' ')
-      end
-
-      def apply_special_formatting(name) # rubocop:disable Metrics/MethodLength
-        name
-          .gsub(/(\d{4}) (\d{2}) (\d{2})/, '\1\2\3')
-          .gsub(/^Gpt /, 'GPT-')
-          .gsub(/^O([13]) /, 'O\1-')
-          .gsub(/^Chatgpt /, 'ChatGPT-')
-          .gsub(/^Tts /, 'TTS-')
-          .gsub(/^Dall E /, 'DALL-E-')
-          .gsub(/3\.5 /, '3.5-')
-          .gsub(/4 /, '4-')
-          .gsub(/4o (?=Mini|Preview|Turbo|Audio)/, '4o-')
-          .gsub(/\bHd\b/, 'HD')
-      end
-    end
-  end
-end