Improve OpenRouter Media Support

Author: sirwolfgang

lib/ruby_llm/providers/anthropic/chat.rb

@@ -89,7 +89,7 @@ def format_message(msg)
         def format_basic_message(msg)
           {
             role: convert_role(msg.role),
-            content: Media.format_content(msg.content)
+            content: self::Media.format_content(msg.content)
           }
         end
 

lib/ruby_llm/providers/bedrock/chat.rb

@@ -29,7 +29,7 @@ def format_message(msg)
         def format_basic_message(msg)
           {
             role: Anthropic::Chat.convert_role(msg.role),
-            content: Media.format_content(msg.content)
+            content: self::Media.format_content(msg.content)
           }
         end
 

lib/ruby_llm/providers/deepseek/media.rb

@@ -0,0 +1,14 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    module DeepSeek
+      # Handles formatting of media content (images, audio) for DeepSeek APIs
+      module Media
+        include OpenAI::Media
+
+        module_function :format_content, :format_image, :format_pdf, :format_audio, :format_text
+      end
+    end
+  end
+end

lib/ruby_llm/providers/gemini/chat.rb

@@ -61,7 +61,7 @@ def format_parts(msg)
               }
             }]
           else
-            Media.format_content(msg.content)
+            self::Media.format_content(msg.content)
           end
         end
 

lib/ruby_llm/providers/openai/chat.rb

@@ -49,7 +49,7 @@ def format_messages(messages)
           messages.map do |msg|
             {
               role: format_role(msg.role),
-              content: Media.format_content(msg.content),
+              content: self::Media.format_content(msg.content),
               tool_calls: format_tool_calls(msg.tool_calls),
               tool_call_id: msg.tool_call_id
             }.compact

lib/ruby_llm/providers/openrouter.rb

@@ -6,6 +6,7 @@ module Providers
     module OpenRouter
       extend OpenAI
       extend OpenRouter::Models
+      extend OpenRouter::Media
 
       module_function
 

lib/ruby_llm/providers/openrouter/media.rb

@@ -0,0 +1,72 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    module OpenRouter
+      # Handles formatting of media content (images, audio) for OpenRouter APIs
+      module Media
+        module_function
+
+        def format_content(content) # rubocop:disable Metrics/MethodLength
+          return content unless content.is_a?(Array)
+
+          content.map do |part|
+            case part[:type]
+            when 'image'
+              format_image(part)
+            when 'input_audio'
+              format_audio(part)
+            when 'pdf'
+              format_pdf(part)
+            else
+              part
+            end
+          end
+        end
+
+        # @see https://openrouter.ai/docs/features/images-and-pdfs#image-inputs
+        def format_image(part)
+          {
+            type: 'image_url',
+            image_url: {
+              url: format_image_url(part[:source]),
+              detail: 'auto'
+            }
+          }
+        end
+
+        def format_image_url(source)
+          if source[:type] == 'base64'
+            "data:#{source[:media_type]};base64,#{source[:data]}"
+          else
+            source[:url]
+          end
+        end
+
+        def format_audio(part)
+          {
+            type: 'input_audio',
+            input_audio: part[:input_audio]
+          }
+        end
+
+        # @see https://openrouter.ai/docs/features/images-and-pdfs#pdf-support
+        def format_pdf(part)
+          {
+            type: 'file',
+            file: {
+              filename: File.basename(part[:source]),
+              file_data: format_file_data(part[:content] || part[:source])
+            }
+          }
+        end
+
+        def format_file_data(source)
+          source = Faraday.get(source).body if source.start_with?('http')
+
+          "data:application/pdf;base64,#{Base64.strict_encode64(source)}"
+        end
+      end
+    end
+  end
+end