feat: add OpenaiMultimodalContentService for content generation

app/services/llm_formatter/conversation_llm_formatter.rb

@@ -5,7 +5,7 @@ class LlmFormatter::ConversationLlmFormatter < LlmFormatter::DefaultLlmFormatter
     sections << "Channel: #{@record.inbox.channel.name}"
     sections << 'Message History:'
     sections << if @record.messages.any?
-                  build_messages
+                  build_messages(config)
                 else
                   'No messages in this conversation'
                 end
@@ -18,11 +18,16 @@ class LlmFormatter::ConversationLlmFormatter < LlmFormatter::DefaultLlmFormatter
 
   private
 
-  def build_messages
+  def build_messages(config = {})
     return "No messages in this conversation\n" if @record.messages.empty?
 
     message_text = ''
-    @record.messages.where.not(message_type: :activity).order(created_at: :asc).each do |message|
+    messages = @record.messages.where.not(message_type: :activity).order(created_at: :asc)
+
+    messages.each do |message|
+      # Skip private messages unless explicitly included in config
+      next if message.private? && !config[:include_private_messages]
+
       message_text << format_message(message)
     end
     message_text
@@ -30,7 +35,7 @@ class LlmFormatter::ConversationLlmFormatter < LlmFormatter::DefaultLlmFormatter
 
   def format_message(message)
     sender = message.message_type == 'incoming' ? 'User' : 'Support agent'
-    sender = "[Private] #{sender}" if message.private?
+    sender = "[Private Note] #{sender}" if message.private?
     "#{sender}: #{message.content}\n"
   end
 

app/services/openai_multimodal_content_service.rb

@@ -0,0 +1,74 @@
+class OpenaiMultimodalContentService
+  def initialize(message)
+    @message = message
+  end
+
+  def generate_content
+    parts = []
+
+    parts << text_part(@message.content) if @message.content.present?
+    parts.concat(attachment_parts(@message.attachments)) if @message.attachments.any?
+
+    finalize_content_parts(parts)
+  end
+
+  private
+
+  def text_part(text)
+    { type: 'text', text: text }
+  end
+
+  def attachment_parts(attachments)
+    [].tap do |parts|
+      parts.concat(image_parts(attachments.where(file_type: :image)))
+
+      transcription = extract_audio_transcriptions(attachments)
+      parts << text_part(transcription) if transcription.present?
+
+      parts << text_part('User has shared an attachment') if attachments.where.not(file_type: %i[image audio]).exists?
+    end
+  end
+
+  def image_parts(image_attachments)
+    image_attachments.each_with_object([]) do |attachment, parts|
+      url = get_attachment_url(attachment)
+      next if url.blank?
+
+      parts << {
+        type: 'image_url',
+        image_url: { url: url }
+      }
+    end
+  end
+
+  def finalize_content_parts(parts)
+    return 'Message without content' if parts.blank?
+    return parts.first[:text] if single_text_part?(parts)
+
+    parts
+  end
+
+  def single_text_part?(parts)
+    parts.one? && parts.first[:type] == 'text'
+  end
+
+  def get_attachment_url(attachment)
+    return attachment.external_url if attachment.external_url.present?
+
+    return unless attachment.file.attached?
+
+    attachment.file_url
+  end
+
+  def extract_audio_transcriptions(attachments)
+    audio_attachments = attachments.where(file_type: :audio)
+    return '' if audio_attachments.blank?
+
+    transcriptions = ''
+    audio_attachments.each do |attachment|
+      result = Messages::AudioTranscriptionService.new(attachment).perform
+      transcriptions += result[:transcriptions] if result[:success]
+    end
+    transcriptions
+  end
+end

enterprise/app/helpers/captain/chat_helper.rb

@@ -18,18 +18,6 @@ module Captain::ChatHelper
     raise e
   end
 
-  def extract_audio_transcriptions(attachments)
-    audio_attachments = attachments.where(file_type: :audio)
-    return '' if audio_attachments.blank?
-
-    transcriptions = ''
-    audio_attachments.each do |attachment|
-      result = Messages::AudioTranscriptionService.new(attachment).perform
-      transcriptions += result[:transcriptions] if result[:success]
-    end
-    transcriptions
-  end
-
   private
 
   def handle_response(response)

enterprise/app/jobs/captain/conversation/response_builder_job.rb

@@ -38,14 +38,12 @@ class Captain::Conversation::ResponseBuilderJob < ApplicationJob
     account.increment_response_usage
   end
 
-  def collect_previous_messages(include_all: false)
-    messages_query = @conversation
-                     .messages
-                     .where(message_type: [:incoming, :outgoing])
-
-    messages_query = messages_query.where(private: false) unless include_all
-
-    messages_query.map do |message|
+  def collect_previous_messages
+    @conversation
+      .messages
+      .where(message_type: [:incoming, :outgoing])
+      .where(private: false)
+      .map do |message|
       {
         content: message_content_multimodal(message),
         role: determine_role(message)
@@ -60,62 +58,7 @@ class Captain::Conversation::ResponseBuilderJob < ApplicationJob
   end
 
   def message_content_multimodal(message)
-    parts = []
-
-    parts << text_part(message.content) if message.content.present?
-    parts.concat(attachment_parts(message.attachments)) if message.attachments.any?
-
-    finalize_content_parts(parts)
-  end
-
-  def text_part(text)
-    { type: 'text', text: text }
-  end
-
-  def attachment_parts(attachments)
-    [].tap do |parts|
-      parts.concat(image_parts(attachments.where(file_type: :image)))
-
-      transcription = extract_audio_transcriptions(attachments)
-      parts << text_part(transcription) if transcription.present?
-
-      parts << text_part('User has shared an attachment') if attachments.where.not(file_type: %i[image audio]).exists?
-    end
-  end
-
-  def image_parts(image_attachments)
-    image_attachments.each_with_object([]) do |attachment, parts|
-      url = get_attachment_url(attachment)
-      next if url.blank?
-
-      parts << {
-        type: 'image_url',
-        image_url: { url: url }
-      }
-    end
-  end
-
-  def finalize_content_parts(parts)
-    return 'Message without content' if parts.blank?
-    return parts.first[:text] if single_text_part?(parts)
-
-    parts
-  end
-
-  def single_text_part?(parts)
-    parts.one? && parts.first[:type] == 'text'
-  end
-
-  def get_attachment_url(attachment)
-    return attachment.external_url if attachment.external_url.present?
-
-    return unless attachment.file.attached?
-
-    begin
-      attachment.file_url
-    rescue ActiveStorage::FileNotFoundError
-      nil
-    end
+    OpenaiMultimodalContentService.new(message).generate_content
   end
 
   def handoff_requested?

enterprise/app/services/captain/tools/copilot/get_conversation_service.rb

@@ -30,7 +30,7 @@ class Captain::Tools::Copilot::GetConversationService < Captain::Tools::BaseServ
     conversation = Conversation.find_by(display_id: conversation_id, account_id: @assistant.account_id)
     return 'Conversation not found' if conversation.blank?
 
-    conversation.to_llm_text
+    conversation.to_llm_text(include_private_messages: true)
   end
 
   def active?

enterprise/lib/chat_gpt.rb

@@ -8,9 +8,9 @@ class ChatGpt
     @messages = [system_message(context_sections)]
   end
 
-  def generate_response(additional_message: nil, message_history: [], role: 'user')
-    @messages += message_history
-    @messages << { 'role': role, 'content': additional_message } if additional_message.present?
+  def generate_response(input, previous_messages = [], role = 'user')
+    @messages += previous_messages
+    @messages << { 'role': role, 'content': input } if input.present?
 
     response = request_gpt
     JSON.parse(response['choices'][0]['message']['content'].strip)