ruby_llm_community 1.0.0 → 1.1.0

data/lib/ruby_llm/models.rb

@@ -155,31 +155,21 @@ module RubyLLM
     end
 
     def load_models
-      # Try to load from database first if configured
-      if RubyLLM.config.model_registry_class
-        load_from_database
-      else
-        load_from_json
-      end
-    rescue StandardError => e
-      RubyLLM.logger.debug "Failed to load models from database: #{e.message}, falling back to JSON"
-      load_from_json
+      read_from_json
     end
 
-    def load_from_database
-      model_class = RubyLLM.config.model_registry_class
-      model_class = model_class.constantize if model_class.is_a?(String)
-      model_class.all.map(&:to_llm)
+    def load_from_json!
+      @models = read_from_json
     end
 
-    def load_from_json
+    def read_from_json
       data = File.exist?(self.class.models_file) ? File.read(self.class.models_file) : '[]'
       JSON.parse(data, symbolize_names: true).map { |model| Model::Info.new(model) }
     rescue JSON::ParserError
       []
     end
 
-    def save_models
+    def save_to_json
       File.write(self.class.models_file, JSON.pretty_generate(all.map(&:to_h)))
     end
 

data/lib/ruby_llm/provider.rb

@@ -42,7 +42,6 @@ module RubyLLM
       normalized_temperature = maybe_normalize_temperature(temperature, model)
 
       payload = Utils.deep_merge(
-        params,
         render_payload(
           messages,
           tools: tools,
@@ -51,7 +50,8 @@ module RubyLLM
           cache_prompts: cache_prompts,
           stream: block_given?,
           schema: schema
-        )
+        ),
+        params
       )
 
       if block_given?
@@ -62,8 +62,10 @@ module RubyLLM
     end
 
     def list_models
-      response = @connection.get models_url
-      parse_list_models_response response, slug, capabilities
+      Array(models_url).flat_map do |url|
+        response = @connection.get(url)
+        parse_list_models_response(response, slug, capabilities)
+      end
     end
 
     def embed(text, model:, dimensions:)

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

@@ -98,7 +98,7 @@ module RubyLLM
           with_cache_control(
             {
               type: 'text',
-              text: Utils.format_text_file_for_llm(text_file)
+              text: text_file.for_llm
             },
             cache:
           )

data/lib/ruby_llm/providers/bedrock/models.rb

@@ -72,7 +72,25 @@ module RubyLLM
           return model_id unless model_data['inferenceTypesSupported']&.include?('INFERENCE_PROFILE')
           return model_id if model_data['inferenceTypesSupported']&.include?('ON_DEMAND')
 
-          "us.#{model_id}"
+          desired_region_prefix = inference_profile_region_prefix
+
+          # Return unchanged if model already has the correct region prefix
+          return model_id if model_id.start_with?("#{desired_region_prefix}.")
+
+          # Remove any existing region prefix (e.g., "us.", "eu.", "ap.")
+          clean_model_id = model_id.sub(/^[a-z]{2}\./, '')
+
+          # Apply the desired region prefix
+          "#{desired_region_prefix}.#{clean_model_id}"
+        end
+
+        def inference_profile_region_prefix
+          # Extract region prefix from bedrock_region (e.g., "eu-west-3" -> "eu")
+          region = @config.bedrock_region.to_s
+          return 'us' if region.empty? # Default fallback
+
+          # Take first two characters as the region prefix
+          region[0, 2]
         end
       end
     end

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

@@ -39,7 +39,7 @@ module RubyLLM
 
         def format_text_file(text_file)
           {
-            text: Utils.format_text_file_for_llm(text_file)
+            text: text_file.for_llm
           }
         end
 

data/lib/ruby_llm/providers/gpustack/media.rb

@@ -34,7 +34,7 @@ module RubyLLM
           {
             type: 'image_url',
             image_url: {
-              url: "data:#{image.mime_type};base64,#{image.encoded}",
+              url: image.for_llm,
               detail: 'auto'
             }
           }

data/lib/ruby_llm/providers/ollama/media.rb

@@ -34,7 +34,7 @@ module RubyLLM
           {
             type: 'image_url',
             image_url: {
-              url: "data:#{image.mime_type};base64,#{image.encoded}",
+              url: image.for_llm,
               detail: 'auto'
             }
           }

data/lib/ruby_llm/providers/openai/media.rb

@@ -36,7 +36,7 @@ module RubyLLM
           {
             type: 'image_url',
             image_url: {
-              url: image.url? ? image.source : "data:#{image.mime_type};base64,#{image.encoded}"
+              url: image.url? ? image.source : image.for_llm
             }
           }
         end
@@ -46,7 +46,7 @@ module RubyLLM
             type: 'file',
             file: {
               filename: pdf.filename,
-              file_data: "data:#{pdf.mime_type};base64,#{pdf.encoded}"
+              file_data: pdf.for_llm
             }
           }
         end
@@ -54,7 +54,7 @@ module RubyLLM
         def format_text_file(text_file)
           {
             type: 'text',
-            text: Utils.format_text_file_for_llm(text_file)
+            text: text_file.for_llm
           }
         end
 
@@ -63,7 +63,7 @@ module RubyLLM
             type: 'input_audio',
             input_audio: {
               data: audio.encoded,
-              format: audio.mime_type.split('/').last
+              format: audio.format
             }
           }
         end

data/lib/ruby_llm/providers/openai/response.rb

@@ -82,7 +82,7 @@ module RubyLLM
 
         def format_image_generation_message(msg)
           items = []
-          image_attachment = msg.content.attachments.first
+          image_attachment = msg.content.attachments.last
           if image_attachment.reasoning_id
             items << {
               type: 'reasoning',
@@ -144,7 +144,7 @@ module RubyLLM
 
           return text_content unless image_outputs.any?
 
-          build_content_with_images(text_content, image_outputs)
+          build_content_with_images(image_outputs, text_content)
         end
 
         private
@@ -157,9 +157,9 @@ module RubyLLM
           end.join("\n")
         end
 
-        def build_content_with_images(text_content, image_outputs)
-          content = RubyLLM::Content.new(text_content)
+        def build_content_with_images(image_outputs, text_content)
           reasoning_id = extract_reasoning_id(@current_outputs)
+          content = RubyLLM::Content.new(text_content)
           image_outputs.each do |output|
             attach_image_to_content(content, output, reasoning_id)
           end
@@ -177,7 +177,8 @@ module RubyLLM
               mime_type: mime_type,
               model_id: nil,
               id: output['id'],
-              reasoning_id: reasoning_id
+              reasoning_id: reasoning_id,
+              revised_prompt: output['revised_prompt']
             )
           )
         end
@@ -194,7 +195,7 @@ module RubyLLM
           msg.role == :assistant &&
             msg.content.is_a?(RubyLLM::Content) &&
             msg.content.attachments.any? &&
-            msg.content.attachments.first.is_a?(RubyLLM::ImageAttachment)
+            msg.content.attachments.last.is_a?(RubyLLM::ImageAttachment)
         end
 
         def extract_reasoning_id(outputs)

data/lib/ruby_llm/providers/openai/response_media.rb

@@ -50,7 +50,7 @@ module RubyLLM
         def format_text_file(text_file)
           {
             type: 'input_text',
-            text: Utils.format_text_file_for_llm(text_file)
+            text: text_file.for_llm
           }
         end
 

data/lib/ruby_llm/providers/openai/streaming.rb

@@ -168,7 +168,7 @@ module RubyLLM
         end
 
         def build_partial_image_chunk(data)
-          content = build_image_content(data['partial_image_b64'], 'image/png', nil, nil)
+          content = build_image_content(item: data, text_content: nil, revised_prompt: nil, partial: true)
 
           Chunk.new(
             role: :assistant,
@@ -182,12 +182,10 @@ module RubyLLM
 
         def build_completed_image_chunk(data)
           item = data['item']
-          image_data = item['result']
-          output_format = item['output_format'] || 'png'
-          mime_type = "image/#{output_format}"
+          text_content = item['delta'] || ''
           revised_prompt = item['revised_prompt']
 
-          content = build_image_content(image_data, mime_type, nil, revised_prompt)
+          content = build_image_content(item:, text_content:, revised_prompt:, partial: false)
 
           Chunk.new(
             role: :assistant,
@@ -211,14 +209,19 @@ module RubyLLM
           )
         end
 
-        def build_image_content(base64_data, mime_type, model_id, revised_prompt = nil)
-          text_content = revised_prompt || ''
-          content = RubyLLM::Content.new(text_content)
+        def build_image_content(item:, text_content:, revised_prompt: nil, partial: false)
+          content = RubyLLM::Content.new(text_content || '')
+          output_format = item['output_format'] || 'png'
+          mime_type = "image/#{output_format}"
+
           content.attach(
             RubyLLM::ImageAttachment.new(
-              data: base64_data,
-              mime_type: mime_type,
-              model_id: model_id
+              data: item['result'] || item['partial_image_b64'],
+              mime_type:,
+              model_id: item['model_id'],
+              id: item['id'],
+              revised_prompt:,
+              partial:
             )
           )
           content

data/lib/ruby_llm/providers/openai/tools.rb

@@ -57,6 +57,16 @@ module RubyLLM
           end
         end
 
+        def parse_tool_call_arguments(tool_call)
+          arguments = tool_call.dig('function', 'arguments')
+
+          if arguments.nil? || arguments.empty?
+            {}
+          else
+            JSON.parse(arguments)
+          end
+        end
+
         def parse_tool_calls(tool_calls, parse_arguments: true)
           return nil unless tool_calls&.any?
 
@@ -67,12 +77,7 @@ module RubyLLM
                 id: tc['id'],
                 name: tc.dig('function', 'name'),
                 arguments: if parse_arguments
-                             if tc.dig('function', 'arguments').empty?
-                               {}
-                             else
-                               JSON.parse(tc.dig('function',
-                                                 'arguments'))
-                             end
+                             parse_tool_call_arguments(tc)
                            else
                              tc.dig('function', 'arguments')
                            end

data/lib/ruby_llm/providers/vertexai.rb

@@ -33,7 +33,7 @@ module RubyLLM
       private
 
       def access_token
-        return 'test-token' if defined?(VCR) && VCR.current_cassette
+        return 'test-token' if defined?(VCR) && !VCR.current_cassette.recording?
 
         initialize_authorizer unless @authorizer
         @authorizer.fetch_access_token!['access_token']

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

@@ -0,0 +1,166 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class XAI
+      # Determines capabilities and pricing for xAI (Grok) models
+      # - https://docs.x.ai/docs/models
+      module Capabilities
+        module_function
+
+        # rubocop:disable Naming/VariableNumber
+        MODEL_PATTERNS = {
+          grok_2: /^grok-2(?!-vision)/,
+          grok_2_vision: /^grok-2-vision/,
+          grok_2_image: /^grok-2-image/,
+          grok_3: /^grok-3(?!-(?:fast|mini))/,
+          grok_3_fast: /^grok-3-fast/,
+          grok_3_mini: /^grok-3-mini(?!-fast)/,
+          grok_3_mini_fast: /^grok-3-mini-fast/,
+          grok_4: /^grok-4/
+        }.freeze
+        # rubocop:enable Naming/VariableNumber
+
+        def context_window_for(model_id)
+          case model_family(model_id)
+          when 'grok_4' then 256_000
+          when 'grok_2_vision' then 32_768
+          else 131_072
+          end
+        end
+
+        def max_tokens_for(_model_id)
+          4_096
+        end
+
+        def supports_vision?(model_id)
+          case model_family(model_id)
+          when 'grok_2_vision' then true
+          else false
+          end
+        end
+
+        def supports_functions?(model_id)
+          model_family(model_id) != 'grok_2_image'
+        end
+
+        def supports_structured_output?(model_id)
+          model_family(model_id) != 'grok_2_image'
+        end
+
+        def supports_json_mode?(model_id)
+          supports_structured_output?(model_id)
+        end
+
+        # Pricing from API data (per million tokens)
+        # rubocop:disable Naming/VariableNumber
+        PRICES = {
+          grok_2: { input: 2.0, output: 10.0 },
+          grok_2_vision: { input: 2.0, output: 10.0 },
+          grok_3: { input: 3.0, output: 15.0, cached_input: 0.75 },
+          grok_3_fast: { input: 5.0, output: 25.0, cached_input: 1.25 },
+          grok_3_mini: { input: 0.3, output: 0.5, cached_input: 0.075 },
+          grok_3_mini_fast: { input: 0.6, output: 4.0, cached_input: 0.15 },
+          grok_4: { input: 3.0, output: 15.0, cached_input: 0.75 }
+        }.freeze
+        # rubocop:enable Naming/VariableNumber
+
+        def model_family(model_id)
+          MODEL_PATTERNS.each do |family, pattern|
+            return family.to_s if model_id.match?(pattern)
+          end
+          'other'
+        end
+
+        def input_price_for(model_id)
+          family = model_family(model_id).to_sym
+          prices = PRICES.fetch(family, { input: default_input_price })
+          prices[:input] || default_input_price
+        end
+
+        def cached_input_price_for(model_id)
+          family = model_family(model_id).to_sym
+          prices = PRICES.fetch(family, {})
+          prices[:cached_input]
+        end
+
+        def output_price_for(model_id)
+          family = model_family(model_id).to_sym
+          prices = PRICES.fetch(family, { output: default_output_price })
+          prices[:output] || default_output_price
+        end
+
+        def model_type(model_id)
+          return 'image' if model_family(model_id) == 'grok_2_image'
+
+          'chat'
+        end
+
+        def default_input_price
+          2.0
+        end
+
+        def default_output_price
+          10.0
+        end
+
+        def format_display_name(model_id)
+          model_id.then { |id| humanize(id) }
+                  .then { |name| apply_special_formatting(name) }
+        end
+
+        def humanize(id)
+          id.tr('-', ' ')
+            .split
+            .map(&:capitalize)
+            .join(' ')
+        end
+
+        def apply_special_formatting(name)
+          name
+            .gsub(/^Grok /, 'Grok-')
+            .gsub(/(\d{4}) (\d{2}) (\d{2})/, '\1-\2-\3')
+        end
+
+        def modalities_for(model_id)
+          modalities = {
+            input: ['text'],
+            output: []
+          }
+
+          modalities[:output] << 'text' if model_type(model_id) == 'chat'
+
+          # Vision support
+          modalities[:input] << 'image' if supports_vision?(model_id)
+
+          modalities
+        end
+
+        def capabilities_for(model_id)
+          capabilities = []
+
+          # Common capabilities
+          capabilities << 'streaming'
+          capabilities << 'function_calling' if supports_functions?(model_id)
+          capabilities << 'structured_output' if supports_structured_output?(model_id)
+
+          capabilities
+        end
+
+        def pricing_for(model_id)
+          standard_pricing = {
+            input_per_million: input_price_for(model_id),
+            output_per_million: output_price_for(model_id)
+          }
+
+          # Add cached pricing if available
+          cached_price = cached_input_price_for(model_id)
+          standard_pricing[:cached_input_per_million] = cached_price if cached_price
+
+          # Pricing structure
+          { text_tokens: { standard: standard_pricing } }
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,15 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class XAI
+      # Chat implementation for xAI
+      # https://docs.x.ai/docs/api-reference#chat-completions
+      module Chat
+        def format_role(role)
+          role.to_s
+        end
+      end
+    end
+  end
+end

data/lib/ruby_llm/providers/xai/models.rb

@@ -0,0 +1,48 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class XAI
+      # Model definitions for xAI API
+      # https://docs.x.ai/docs/api-reference#list-language-models
+      # https://docs.x.ai/docs/api-reference#list-image-generation-models
+      module Models
+        module_function
+
+        # NOTE: We pull models list from two endpoints here as these provide
+        #       detailed, modality, capability and cost information for each
+        #       model that we can leverage which the generic OpenAI compatible
+        #       /models endpoint does not provide.
+        def models_url
+          %w[language-models image-generation-models]
+        end
+
+        def parse_list_models_response(response, slug, capabilities)
+          data = response.body
+          return [] if data.empty?
+
+          data['models']&.map do |model_data|
+            model_id = model_data['id']
+
+            Model::Info.new(
+              id: model_id,
+              name: capabilities.format_display_name(model_id),
+              provider: slug,
+              family: capabilities.model_family(model_id),
+              modalities: {
+                input: model_data['input_modalities'] | capabilities.modalities_for(model_id)[:input],
+                output: model_data['output_modalities'] | capabilities.modalities_for(model_id)[:output]
+              },
+              context_window: capabilities.context_window_for(model_id),
+              capabilities: capabilities.capabilities_for(model_id),
+              pricing: capabilities.pricing_for(model_id),
+              metadata: {
+                aliases: model_data['aliases']
+              }
+            )
+          end || []
+        end
+      end
+    end
+  end
+end

data/lib/ruby_llm/providers/xai.rb

@@ -0,0 +1,46 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    # xAI API integration
+    class XAI < OpenAIBase
+      include XAI::Capabilities
+      include XAI::Chat
+      include XAI::Models
+
+      def api_base
+        'https://api.x.ai/v1'
+      end
+
+      def headers
+        {
+          'Authorization' => "Bearer #{@config.xai_api_key}",
+          'Content-Type' => 'application/json'
+        }
+      end
+
+      # xAI uses a different error format than OpenAI
+      # {"code": "...", "error": "..."}
+      def parse_error(response)
+        return if response.body.empty?
+
+        body = try_parse_json(response.body)
+        case body
+        when Hash then body['error']
+        when Array then body.map { |part| part['error'] }.join('. ')
+        else body
+        end
+      end
+
+      class << self
+        def capabilities
+          XAI::Capabilities
+        end
+
+        def configuration_requirements
+          %i[xai_api_key]
+        end
+      end
+    end
+  end
+end

data/lib/ruby_llm/railtie.rb

@@ -3,14 +3,31 @@
 module RubyLLM
   # Rails integration for RubyLLM
   class Railtie < Rails::Railtie
+    initializer 'ruby_llm.inflections' do
+      ActiveSupport::Inflector.inflections(:en) do |inflect|
+        inflect.acronym 'LLM'
+      end
+    end
+
     initializer 'ruby_llm.active_record' do
       ActiveSupport.on_load :active_record do
-        include RubyLLM::ActiveRecord::ActsAs
+        if RubyLLM.config.use_new_acts_as
+          require 'ruby_llm/active_record/acts_as'
+          ::ActiveRecord::Base.include RubyLLM::ActiveRecord::ActsAs
+        else
+          require 'ruby_llm/active_record/acts_as_legacy'
+          ::ActiveRecord::Base.include RubyLLM::ActiveRecord::ActsAsLegacy
+
+          Rails.logger.warn(
+            "\n!!! RubyLLM's legacy acts_as API is deprecated and will be removed in RubyLLM 2.0.0. " \
+            "Please consult the migration guide at https://rubyllm.com/upgrading-to-1-7/\n"
+          )
+        end
       end
     end
 
-    generators do
-      require 'generators/ruby_llm/install_generator'
+    rake_tasks do
+      load 'tasks/ruby_llm.rake'
     end
   end
 end

data/lib/ruby_llm/stream_accumulator.rb

@@ -72,11 +72,9 @@ module RubyLLM
         when [String, String]
           @content << new_content
         when [String, Content]
-          # Convert accumulated string to Content and merge
           @content = Content.new(@content)
           merge_content(new_content)
         when [Content, String]
-          # Append string to existing Content's text
           @content.instance_variable_set(:@text, (@content.text || '') + new_content)
         when [Content, Content]
           merge_content(new_content)
@@ -85,12 +83,10 @@ module RubyLLM
     end
 
     def merge_content(new_content)
-      # Merge text
       current_text = @content.text || ''
       new_text = new_content.text || ''
       @content.instance_variable_set(:@text, current_text + new_text)
 
-      # Merge attachments
       new_content.attachments.each do |attachment|
         @content.attach(attachment)
       end