ruby_llm 0.1.0.pre35 → 0.1.0.pre37

data/lib/ruby_llm/providers/anthropic/capabilities.rb

@@ -7,45 +7,73 @@ module RubyLLM
       module Capabilities
         module_function
 
-        def determine_context_window(model_id)
-          case model_id
-          when /claude-3/ then 200_000
-          else 100_000
-          end
+        # Determines the context window size for a given model
+        # @param model_id [String] the model identifier
+        # @return [Integer] the context window size in tokens
+        def determine_context_window(_model_id)
+          # All Claude 3 and 3.5 and 3.7 models have 200K token context windows
+          200_000
         end
 
+        # Determines the maximum output tokens for a given model
+        # @param model_id [String] the model identifier
+        # @return [Integer] the maximum output tokens
         def determine_max_tokens(model_id)
           case model_id
-          when /claude-3-5/ then 8_192
-          else 4_096
+          when /claude-3-(7-sonnet|5)/ then 8_192 # Can be increased to 64K with extended thinking
+          else 4_096 # Claude 3 Opus and Haiku
           end
         end
 
+        # Gets the input price per million tokens for a given model
+        # @param model_id [String] the model identifier
+        # @return [Float] the price per million tokens for input
         def get_input_price(model_id)
           PRICES.dig(model_family(model_id), :input) || default_input_price
         end
 
+        # Gets the output price per million tokens for a given model
+        # @param model_id [String] the model identifier
+        # @return [Float] the price per million tokens for output
         def get_output_price(model_id)
           PRICES.dig(model_family(model_id), :output) || default_output_price
         end
 
+        # Determines if a model supports vision capabilities
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports vision
         def supports_vision?(model_id)
-          return false if model_id.match?(/claude-3-5-haiku/)
-          return false if model_id.match?(/claude-[12]/)
-
-          true
+          # All Claude 3, 3.5, and 3.7 models support vision
+          !model_id.match?(/claude-[12]/)
         end
 
+        # Determines if a model supports function calling
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports functions
         def supports_functions?(model_id)
-          model_id.include?('claude-3')
+          model_id.match?(/claude-3/)
         end
 
+        # Determines if a 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.include?('claude-3')
+          model_id.match?(/claude-3/)
+        end
+
+        # Determines if a model supports extended thinking
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports extended thinking
+        def supports_extended_thinking?(model_id)
+          model_id.match?(/claude-3-7-sonnet/)
         end
 
+        # Determines the model family for a given model ID
+        # @param model_id [String] the model identifier
+        # @return [Symbol] the model family identifier
         def model_family(model_id)
           case model_id
+          when /claude-3-7-sonnet/  then :claude37_sonnet
           when /claude-3-5-sonnet/  then :claude35_sonnet
           when /claude-3-5-haiku/   then :claude35_haiku
           when /claude-3-opus/      then :claude3_opus
@@ -55,23 +83,32 @@ module RubyLLM
           end
         end
 
+        # Returns the model type
+        # @param model_id [String] the model identifier (unused but kept for API consistency)
+        # @return [String] the model type, always 'chat' for Anthropic models
         def model_type(_)
           'chat'
         end
 
+        # Pricing information for Anthropic models (per million tokens)
         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
+          claude37_sonnet: { input: 3.0, output: 15.0 },   # $3.00/$15.00 per million tokens
+          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
 
+        # Default input price if model not found in PRICES
+        # @return [Float] default price per million tokens for input
         def default_input_price
           3.0
         end
 
+        # Default output price if model not found in PRICES
+        # @return [Float] default price per million tokens for output
         def default_output_price
           15.0
         end

data/lib/ruby_llm/providers/anthropic/chat.rb

@@ -35,7 +35,7 @@ module RubyLLM
 
         def extract_text_content(blocks)
           text_blocks = blocks.select { |c| c['type'] == 'text' }
-          text_blocks.map { |c| c['text'] }.join('')
+          text_blocks.map { |c| c['text'] }.join
         end
 
         def build_message(data, content, tool_use)
@@ -68,8 +68,7 @@ module RubyLLM
 
         def convert_role(role)
           case role
-          when :tool then 'user'
-          when :user then 'user'
+          when :tool, :user then 'user'
           else 'assistant'
           end
         end

data/lib/ruby_llm/providers/deepseek/capabilities.rb

@@ -7,6 +7,9 @@ module RubyLLM
       module Capabilities
         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
@@ -14,6 +17,9 @@ module RubyLLM
           end
         end
 
+        # Returns the maximum number of tokens that can be generated
+        # @param model_id [String] the model identifier
+        # @return [Integer] the maximum number of tokens
         def max_tokens_for(model_id)
           case model_id
           when /deepseek-(?:chat|reasoner)/ then 8_192
@@ -21,30 +27,51 @@ module RubyLLM
           end
         end
 
+        # Returns the price per million tokens for input (cache miss)
+        # @param model_id [String] the model identifier
+        # @return [Float] the price per million tokens in USD
         def input_price_for(model_id)
           PRICES.dig(model_family(model_id), :input_miss) || default_input_price
         end
 
+        # Returns the price per million tokens for output
+        # @param model_id [String] the model identifier
+        # @return [Float] the price per million tokens in USD
         def output_price_for(model_id)
           PRICES.dig(model_family(model_id), :output) || default_output_price
         end
 
+        # Returns the price per million tokens for input with cache hit
+        # @param model_id [String] the model identifier
+        # @return [Float] the price per million tokens in USD
         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
 
+        # Returns a formatted display name for the model
+        # @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'
@@ -56,13 +83,18 @@ module RubyLLM
           end
         end
 
+        # Returns the model type
+        # @param model_id [String] the model identifier
+        # @return [String] the model type (e.g., 'chat')
         def model_type(_model_id)
           'chat' # All DeepSeek models are chat models
         end
 
+        # Returns the model family
+        # @param model_id [String] the model identifier
+        # @return [Symbol] the model family
         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
@@ -84,14 +116,20 @@ module RubyLLM
 
         private
 
+        # Default input price when model family can't be determined
+        # @return [Float] the default input price
         def default_input_price
           0.27 # Default to chat cache miss price
         end
 
+        # Default output price when model family can't be determined
+        # @return [Float] the default output price
         def default_output_price
           1.10 # Default to chat output price
         end
 
+        # Default cache hit price when model family can't be determined
+        # @return [Float] the default cache hit price
         def default_cache_hit_price
           0.07 # Default to chat cache hit price
         end

data/lib/ruby_llm/providers/gemini/capabilities.rb

@@ -7,25 +7,34 @@ module RubyLLM
       module Capabilities # rubocop:disable Metrics/ModuleLength
         module_function
 
+        # Returns the context window size (input token limit) 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 /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 /text-embedding-004/, /embedding-001/ then 2_048
           when /aqa/ then 7_168
           else 32_768 # Sensible default for unknown models
           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 /gemini-2\.0-flash/, /gemini-1\.5/ then 8_192
-          when /text-embedding/, /embedding-001/ then 768 # Output dimension size for embeddings
+          when /text-embedding-004/, /embedding-001/ then 768 # Output dimension size for embeddings
           when /aqa/ then 1_024
           else 4_096 # Sensible default
           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 in USD
         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)
@@ -34,6 +43,9 @@ module RubyLLM
           context_length(model_id) > 128_000 ? base_price * 2 : base_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 in USD
         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)
@@ -42,6 +54,9 @@ module RubyLLM
           context_length(model_id) > 128_000 ? base_price * 2 : base_price
         end
 
+        # Determines if the model supports vision (image/video) inputs
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports vision inputs
         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/)
@@ -49,6 +64,9 @@ module RubyLLM
           model_id.match?(/gemini-[12]\.[05]/)
         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)
           return false if model_id.match?(/text-embedding|embedding-001|aqa/)
           return false if model_id.match?(/flash-lite/)
@@ -57,13 +75,20 @@ module RubyLLM
           model_id.match?(/gemini-[12]\.[05]-(?:pro|flash)(?!-lite)/)
         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)
           return false if model_id.match?(/text-embedding|embedding-001|aqa/)
           return false if model_id.match?(/gemini-1\.0/)
+          return false if model_id.match?(/gemini-2\.0-flash-lite/)
 
           model_id.match?(/gemini-\d/)
         end
 
+        # Formats the model ID into a human-readable display name
+        # @param model_id [String] the model identifier
+        # @return [String] the formatted display name
         def format_display_name(model_id)
           model_id
             .delete_prefix('models/')
@@ -72,24 +97,36 @@ module RubyLLM
             .join(' ')
             .gsub(/(\d+\.\d+)/, ' \1') # Add space before version numbers
             .gsub(/\s+/, ' ')          # Clean up multiple spaces
-            .gsub(/Aqa/, 'AQA')        # Special case for AQA
+            .gsub('Aqa', 'AQA')        # Special case for AQA
             .strip
         end
 
+        # Determines if the model supports context caching
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports caching
         def supports_caching?(model_id)
           return false if model_id.match?(/flash-lite|gemini-1\.0/)
 
           model_id.match?(/gemini-[12]\.[05]/)
         end
 
+        # Determines if the model supports tuning
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports tuning
         def supports_tuning?(model_id)
           model_id.match?(/gemini-1\.5-flash/)
         end
 
+        # Determines if the model supports audio inputs
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports audio inputs
         def supports_audio?(model_id)
           model_id.match?(/gemini-[12]\.[05]/)
         end
 
+        # Returns the type of model (chat, embedding, image)
+        # @param model_id [String] the model identifier
+        # @return [String] the model type
         def model_type(model_id)
           case model_id
           when /text-embedding|embedding/ then 'embedding'
@@ -98,6 +135,9 @@ module RubyLLM
           end
         end
 
+        # Returns the model family identifier
+        # @param model_id [String] the model identifier
+        # @return [String] the model family identifier
         def model_family(model_id) # rubocop:disable Metrics/CyclomaticComplexity,Metrics/MethodLength
           case model_id
           when /gemini-2\.0-flash-lite/ then 'gemini20_flash_lite'
@@ -113,7 +153,10 @@ module RubyLLM
           end
         end
 
-        def pricing_family(model_id) # rubocop:disable Metrics/CyclomaticComplexity
+        # Returns the pricing family identifier for the model
+        # @param model_id [String] the model identifier
+        # @return [Symbol] the pricing family identifier
+        def pricing_family(model_id) # rubocop:disable Metrics/CyclomaticComplexity,Metrics/MethodLength
           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
@@ -122,18 +165,26 @@ module RubyLLM
           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
+          when /aqa/ then :aqa
           else :base
           end
         end
 
+        # Determines if the model supports long context
+        # @param model_id [String] the model identifier
+        # @return [Boolean] true if the model supports long context
         def long_context_model?(model_id)
           model_id.match?(/gemini-1\.5-(?:pro|flash)/)
         end
 
+        # Returns the context length for the model
+        # @param model_id [String] the model identifier
+        # @return [Integer] the context length in tokens
         def context_length(model_id)
           context_window_for(model_id)
         end
 
+        # Pricing information for Gemini models (per 1M tokens in USD)
         PRICES = {
           flash_2: { # Gemini 2.0 Flash # rubocop:disable Naming/VariableNumber
             input: 0.10,
@@ -152,19 +203,22 @@ module RubyLLM
             input: 0.075,
             output: 0.30,
             cache: 0.01875,
-            cache_storage: 1.00
+            cache_storage: 1.00,
+            grounding_search: 35.00 # per 1K requests
           },
           flash_8b: { # Gemini 1.5 Flash 8B
             input: 0.0375,
             output: 0.15,
             cache: 0.01,
-            cache_storage: 0.25
+            cache_storage: 0.25,
+            grounding_search: 35.00 # per 1K requests
           },
           pro: { # Gemini 1.5 Pro
             input: 1.25,
             output: 5.0,
             cache: 0.3125,
-            cache_storage: 4.50
+            cache_storage: 4.50,
+            grounding_search: 35.00 # per 1K requests
           },
           pro_1_0: { # Gemini 1.0 Pro # rubocop:disable Naming/VariableNumber
             input: 0.50,
@@ -173,15 +227,23 @@ module RubyLLM
           embedding: { # Text Embedding models
             input: 0.00,
             output: 0.00
+          },
+          aqa: { # AQA model
+            input: 0.00,
+            output: 0.00
           }
         }.freeze
 
+        # Default input price for unknown models
+        # @return [Float] the default input price per million tokens
         def default_input_price
           0.075 # Default to Flash pricing
         end
 
+        # Default output price for unknown models
+        # @return [Float] the default output price per million tokens
         def default_output_price
-          0.30  # Default to Flash pricing
+          0.30 # Default to Flash pricing
         end
       end
     end