Fix inlineData and text not handled correctly for Gemini

lib/ruby_llm/providers/gemini/chat.rb

@@ -112,7 +112,7 @@ def parse_completion_response(response)
 
           Message.new(
             role: :assistant,
-            content: extract_text_parts(parts) || parse_content(data),
+            content: parse_content(data),
             thinking: Thinking.build(
               text: extract_thought_parts(parts),
               signature: extract_thought_signature(parts)
@@ -140,21 +140,13 @@ def parse_content(data)
           candidate = data.dig('candidates', 0)
           return '' unless candidate
 
-          return '' if function_call?(candidate)
-
           parts = candidate.dig('content', 'parts')
           return '' unless parts&.any?
 
-          non_thought_parts = parts.reject { |part| part['thought'] }
-          return '' unless non_thought_parts.any?
-
-          build_response_content(non_thought_parts)
-        end
+          content_parts = parts.reject { |part| part['thought'] || part['functionCall'] }
+          return '' unless content_parts.any?
 
-        def extract_text_parts(parts)
-          text_parts = parts.reject { |p| p['thought'] }
-          content = text_parts.filter_map { |p| p['text'] }.join
-          content.empty? ? nil : content
+          build_response_content(content_parts)
         end
 
         def extract_thought_parts(parts)

lib/ruby_llm/providers/gemini/media.rb

@@ -71,7 +71,11 @@ def build_response_content(parts) # rubocop:disable Metrics/PerceivedComplexity
         text = nil if text.empty?
         return text if attachments.empty?
 
-        Content.new(text:, attachments:)
+        Content.new(text).tap do |content|
+          attachments.each do |attachment|
+            content.add_attachment(attachment.source, filename: attachment.filename)
+          end
+        end
       end
 
       def build_inline_attachment(inline_data, index)

spec/ruby_llm/providers/gemini/chat_spec.rb

@@ -600,6 +600,49 @@
       expect(message.output_tokens).to eq(8)
       expect(message.cached_tokens).to eq(21)
     end
+
+    it 'handles message where both text and attachments are present' do
+      raw_data = 'fake-image-bytes'
+      encoded_data = Base64.strict_encode64(raw_data)
+      response = Struct.new(:body, :env).new(
+        {
+          'candidates' => [
+            {
+              'content' => {
+                'parts' => [
+                  {
+                    'functionCall' => {
+                      'name' => 'lookup_weather',
+                      'args' => { 'city' => 'Paris' }
+                    }
+                  },
+                  { 'text' => 'Here is the result with an image.' },
+                  {
+                    'inlineData' => {
+                      'mimeType' => 'image/png',
+                      'data' => encoded_data
+                    }
+                  }
+                ]
+              }
+            }
+          ],
+          'usageMetadata' => {}
+        },
+        Struct.new(:url).new(Struct.new(:path).new('/v1/models/gemini-2.5-flash-image:generateContent'))
+      )
+
+      provider = RubyLLM::Providers::Gemini.new(RubyLLM.config)
+      message = provider.send(:parse_completion_response, response)
+      attachment = message.content.attachments.first
+
+      expect(message.content).to be_a(RubyLLM::Content)
+      expect(message.content.text).to eq('Here is the result with an image.')
+      expect(message.content.attachments.size).to eq(1)
+      expect(attachment).to be_a(RubyLLM::Attachment)
+      expect(attachment.mime_type).to eq('image/png')
+      expect(attachment.content).to eq(raw_data)
+    end
   end
 
   it 'correctly sums candidatesTokenCount and thoughtsTokenCount' do