feat: Add the support for images in Captain (#11850)

spec/enterprise/controllers/api/v1/accounts/captain/assistants_controller_spec.rb

@@ -211,8 +211,8 @@ RSpec.describe 'Api::V1::Accounts::Captain::Assistants', type: :request do
 
         expect(response).to have_http_status(:success)
         expect(chat_service).to have_received(:generate_response).with(
-          valid_params[:message_content],
-          valid_params[:message_history]
+          additional_message: valid_params[:message_content],
+          message_history: valid_params[:message_history]
         )
         expect(json_response[:content]).to eq('Assistant response')
       end
@@ -232,8 +232,8 @@ RSpec.describe 'Api::V1::Accounts::Captain::Assistants', type: :request do
 
         expect(response).to have_http_status(:success)
         expect(chat_service).to have_received(:generate_response).with(
-          params_without_history[:message_content],
-          []
+          additional_message: params_without_history[:message_content],
+          message_history: []
         )
       end
     end

spec/enterprise/jobs/captain/conversation/response_builder_job_spec.rb

@@ -30,5 +30,142 @@ RSpec.describe Captain::Conversation::ResponseBuilderJob, type: :job do
       account.reload
       expect(account.usage_limits[:captain][:responses][:consumed]).to eq(1)
     end
+
+    context 'when message contains an image' do
+      let(:message_with_image) { create(:message, conversation: conversation, message_type: :incoming, content: 'Can you help with this error?') }
+      let(:image_attachment) { message_with_image.attachments.create!(account: account, file_type: :image, external_url: 'https://example.com/error.jpg') }
+
+      before do
+        image_attachment
+      end
+
+      it 'includes image URL directly in the message content for OpenAI vision analysis' do
+        # Expect the generate_response to receive multimodal content with image URL
+        expect(mock_llm_chat_service).to receive(:generate_response) do |**kwargs|
+          history = kwargs[:message_history]
+          last_entry = history.last
+          expect(last_entry[:content]).to be_an(Array)
+          expect(last_entry[:content].any? { |part| part[:type] == 'text' && part[:text] == 'Can you help with this error?' }).to be true
+          expect(last_entry[:content].any? do |part|
+            part[:type] == 'image_url' && part[:image_url][:url] == 'https://example.com/error.jpg'
+          end).to be true
+          { 'response' => 'I can see the error in your image. It appears to be a database connection issue.' }
+        end
+
+        described_class.perform_now(conversation, assistant)
+      end
+    end
+  end
+
+  describe 'retry mechanisms for image processing' do
+    let(:conversation) { create(:conversation, inbox: inbox, account: account) }
+    let(:mock_llm_chat_service) { instance_double(Captain::Llm::AssistantChatService) }
+    let(:mock_message_builder) { instance_double(Captain::OpenAiMessageBuilderService) }
+
+    before do
+      create(:message, conversation: conversation, content: 'Hello with image', message_type: :incoming)
+      allow(Captain::Llm::AssistantChatService).to receive(:new).and_return(mock_llm_chat_service)
+      allow(Captain::OpenAiMessageBuilderService).to receive(:new).with(message: anything).and_return(mock_message_builder)
+      allow(mock_message_builder).to receive(:generate_content).and_return('Hello with image')
+      allow(mock_llm_chat_service).to receive(:generate_response).and_return({ 'response' => 'Test response' })
+    end
+
+    context 'when ActiveStorage::FileNotFoundError occurs' do
+      it 'handles file errors and triggers handoff' do
+        allow(mock_message_builder).to receive(:generate_content)
+          .and_raise(ActiveStorage::FileNotFoundError, 'Image file not found')
+
+        # For retryable errors, the job should handle them and proceed with handoff
+        described_class.perform_now(conversation, assistant)
+
+        # Verify handoff occurred due to repeated failures
+        expect(conversation.reload.status).to eq('open')
+      end
+
+      it 'succeeds when no error occurs' do
+        # Don't raise any error, should succeed normally
+        allow(mock_message_builder).to receive(:generate_content)
+          .and_return('Image content processed successfully')
+
+        described_class.perform_now(conversation, assistant)
+
+        expect(conversation.messages.outgoing.count).to eq(1)
+        expect(conversation.messages.outgoing.last.content).to eq('Test response')
+      end
+    end
+
+    context 'when Faraday::BadRequestError occurs' do
+      it 'handles API errors and triggers handoff' do
+        allow(mock_llm_chat_service).to receive(:generate_response)
+          .and_raise(Faraday::BadRequestError, 'Bad request to image service')
+
+        described_class.perform_now(conversation, assistant)
+        expect(conversation.reload.status).to eq('open')
+      end
+
+      it 'succeeds when no error occurs' do
+        # Don't raise any error, should succeed normally
+        allow(mock_llm_chat_service).to receive(:generate_response)
+          .and_return({ 'response' => 'Response after retry' })
+
+        described_class.perform_now(conversation, assistant)
+
+        expect(conversation.messages.outgoing.last.content).to eq('Response after retry')
+      end
+    end
+
+    context 'when image processing fails permanently' do
+      before do
+        allow(mock_message_builder).to receive(:generate_content)
+          .and_raise(ActiveStorage::FileNotFoundError, 'Image permanently unavailable')
+      end
+
+      it 'triggers handoff after max retries' do
+        # Since perform_now re-raises retryable errors, simulate the final failure after retries
+        allow(mock_message_builder).to receive(:generate_content)
+          .and_raise(StandardError, 'Max retries exceeded')
+
+        expect(ChatwootExceptionTracker).to receive(:new).and_call_original
+
+        described_class.perform_now(conversation, assistant)
+
+        expect(conversation.reload.status).to eq('open')
+      end
+    end
+
+    context 'when non-retryable error occurs' do
+      let(:standard_error) { StandardError.new('Generic error') }
+
+      before do
+        allow(mock_llm_chat_service).to receive(:generate_response).and_raise(standard_error)
+      end
+
+      it 'handles error and triggers handoff' do
+        expect(ChatwootExceptionTracker).to receive(:new)
+          .with(standard_error, account: account)
+          .and_call_original
+
+        described_class.perform_now(conversation, assistant)
+
+        expect(conversation.reload.status).to eq('open')
+      end
+
+      it 'ensures Current.executed_by is reset' do
+        expect(Current).to receive(:executed_by=).with(assistant)
+        expect(Current).to receive(:executed_by=).with(nil)
+
+        described_class.perform_now(conversation, assistant)
+      end
+    end
+  end
+
+  describe 'job configuration' do
+    it 'has retry_on configuration for retryable errors' do
+      expect(described_class).to respond_to(:retry_on)
+    end
+
+    it 'defines MAX_MESSAGE_LENGTH constant' do
+      expect(described_class::MAX_MESSAGE_LENGTH).to eq(10_000)
+    end
   end
 end

spec/enterprise/services/captain/open_ai_message_builder_service_spec.rb

@@ -0,0 +1,310 @@
+require 'rails_helper'
+
+RSpec.describe Captain::OpenAiMessageBuilderService do
+  subject(:service) { described_class.new(message: message) }
+
+  let(:message) { create(:message, content: 'Hello world') }
+
+  describe '#generate_content' do
+    context 'when message has only text content' do
+      it 'returns the text content directly' do
+        expect(service.generate_content).to eq('Hello world')
+      end
+    end
+
+    context 'when message has no content and no attachments' do
+      let(:message) { create(:message, content: nil) }
+
+      it 'returns default message' do
+        expect(service.generate_content).to eq('Message without content')
+      end
+    end
+
+    context 'when message has text content and attachments' do
+      before do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :image, external_url: 'https://example.com/image.jpg')
+        attachment.save!
+      end
+
+      it 'returns an array of content parts' do
+        result = service.generate_content
+        expect(result).to be_an(Array)
+        expect(result).to include({ type: 'text', text: 'Hello world' })
+        expect(result).to include({ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } })
+      end
+    end
+
+    context 'when message has only non-text attachments' do
+      let(:message) { create(:message, content: nil) }
+
+      before do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :image, external_url: 'https://example.com/image.jpg')
+        attachment.save!
+      end
+
+      it 'returns an array of content parts without text' do
+        result = service.generate_content
+        expect(result).to be_an(Array)
+        expect(result).to include({ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } })
+        expect(result).not_to include(hash_including(type: 'text', text: 'Hello world'))
+      end
+    end
+  end
+
+  describe '#attachment_parts' do
+    let(:message) { create(:message, content: nil) }
+    let(:attachments) { message.attachments }
+
+    context 'with image attachments' do
+      before do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :image, external_url: 'https://example.com/image.jpg')
+        attachment.save!
+      end
+
+      it 'includes image parts' do
+        result = service.send(:attachment_parts, attachments)
+        expect(result).to include({ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } })
+      end
+    end
+
+    context 'with audio attachments' do
+      let(:audio_attachment) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :audio)
+        attachment.save!
+        attachment
+      end
+
+      before do
+        allow(Messages::AudioTranscriptionService).to receive(:new).with(audio_attachment).and_return(
+          instance_double(Messages::AudioTranscriptionService, perform: { success: true, transcriptions: 'Audio transcription text' })
+        )
+      end
+
+      it 'includes transcription text part' do
+        audio_attachment # trigger creation
+        result = service.send(:attachment_parts, attachments)
+        expect(result).to include({ type: 'text', text: 'Audio transcription text' })
+      end
+    end
+
+    context 'with other file types' do
+      before do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :file)
+        attachment.save!
+      end
+
+      it 'includes generic attachment message' do
+        result = service.send(:attachment_parts, attachments)
+        expect(result).to include({ type: 'text', text: 'User has shared an attachment' })
+      end
+    end
+
+    context 'with mixed attachment types' do
+      let(:image_attachment) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :image, external_url: 'https://example.com/image.jpg')
+        attachment.save!
+        attachment
+      end
+
+      let(:audio_attachment) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :audio)
+        attachment.save!
+        attachment
+      end
+
+      let(:document_attachment) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :file)
+        attachment.save!
+        attachment
+      end
+
+      before do
+        allow(Messages::AudioTranscriptionService).to receive(:new).with(audio_attachment).and_return(
+          instance_double(Messages::AudioTranscriptionService, perform: { success: true, transcriptions: 'Audio text' })
+        )
+      end
+
+      it 'includes all relevant parts' do
+        image_attachment    # trigger creation
+        audio_attachment    # trigger creation
+        document_attachment # trigger creation
+
+        result = service.send(:attachment_parts, attachments)
+        expect(result).to include({ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } })
+        expect(result).to include({ type: 'text', text: 'Audio text' })
+        expect(result).to include({ type: 'text', text: 'User has shared an attachment' })
+      end
+    end
+  end
+
+  describe '#image_parts' do
+    let(:message) { create(:message, content: nil) }
+
+    context 'with valid image attachments' do
+      let(:image1) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :image, external_url: 'https://example.com/image1.jpg')
+        attachment.save!
+        attachment
+      end
+
+      let(:image2) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :image, external_url: 'https://example.com/image2.jpg')
+        attachment.save!
+        attachment
+      end
+
+      it 'returns image parts for all valid images' do
+        image1 # trigger creation
+        image2 # trigger creation
+
+        image_attachments = message.attachments.where(file_type: :image)
+        result = service.send(:image_parts, image_attachments)
+
+        expect(result).to include({ type: 'image_url', image_url: { url: 'https://example.com/image1.jpg' } })
+        expect(result).to include({ type: 'image_url', image_url: { url: 'https://example.com/image2.jpg' } })
+      end
+    end
+
+    context 'with image attachments without URLs' do
+      let(:image_attachment) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :image, external_url: nil)
+        attachment.save!
+        attachment
+      end
+
+      before do
+        allow(image_attachment).to receive(:file).and_return(instance_double(ActiveStorage::Attached::One, attached?: false))
+      end
+
+      it 'skips images without valid URLs' do
+        image_attachment # trigger creation
+
+        image_attachments = message.attachments.where(file_type: :image)
+        result = service.send(:image_parts, image_attachments)
+
+        expect(result).to be_empty
+      end
+    end
+  end
+
+  describe '#get_attachment_url' do
+    let(:attachment) do
+      attachment = message.attachments.build(account_id: message.account_id, file_type: :image)
+      attachment.save!
+      attachment
+    end
+
+    context 'when attachment has external_url' do
+      before { attachment.update(external_url: 'https://example.com/image.jpg') }
+
+      it 'returns external_url' do
+        expect(service.send(:get_attachment_url, attachment)).to eq('https://example.com/image.jpg')
+      end
+    end
+
+    context 'when attachment has attached file' do
+      before do
+        attachment.update(external_url: nil)
+        allow(attachment).to receive(:file).and_return(instance_double(ActiveStorage::Attached::One, attached?: true))
+        allow(attachment).to receive(:file_url).and_return('https://local.com/file.jpg')
+        allow(attachment).to receive(:download_url).and_return('')
+      end
+
+      it 'returns file_url' do
+        expect(service.send(:get_attachment_url, attachment)).to eq('https://local.com/file.jpg')
+      end
+    end
+
+    context 'when attachment has no URL or file' do
+      before do
+        attachment.update(external_url: nil)
+        allow(attachment).to receive(:file).and_return(instance_double(ActiveStorage::Attached::One, attached?: false))
+      end
+
+      it 'returns nil' do
+        expect(service.send(:get_attachment_url, attachment)).to be_nil
+      end
+    end
+  end
+
+  describe '#extract_audio_transcriptions' do
+    let(:message) { create(:message, content: nil) }
+
+    context 'with no audio attachments' do
+      it 'returns empty string' do
+        result = service.send(:extract_audio_transcriptions, message.attachments)
+        expect(result).to eq('')
+      end
+    end
+
+    context 'with successful audio transcriptions' do
+      let(:audio1) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :audio)
+        attachment.save!
+        attachment
+      end
+
+      let(:audio2) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :audio)
+        attachment.save!
+        attachment
+      end
+
+      before do
+        allow(Messages::AudioTranscriptionService).to receive(:new).with(audio1).and_return(
+          instance_double(Messages::AudioTranscriptionService, perform: { success: true, transcriptions: 'First audio text. ' })
+        )
+        allow(Messages::AudioTranscriptionService).to receive(:new).with(audio2).and_return(
+          instance_double(Messages::AudioTranscriptionService, perform: { success: true, transcriptions: 'Second audio text.' })
+        )
+      end
+
+      it 'concatenates all successful transcriptions' do
+        audio1 # trigger creation
+        audio2 # trigger creation
+
+        attachments = message.attachments
+        result = service.send(:extract_audio_transcriptions, attachments)
+        expect(result).to eq('First audio text. Second audio text.')
+      end
+    end
+
+    context 'with failed audio transcriptions' do
+      let(:audio_attachment) do
+        attachment = message.attachments.build(account_id: message.account_id, file_type: :audio)
+        attachment.save!
+        attachment
+      end
+
+      before do
+        allow(Messages::AudioTranscriptionService).to receive(:new).with(audio_attachment).and_return(
+          instance_double(Messages::AudioTranscriptionService, perform: { success: false, transcriptions: nil })
+        )
+      end
+
+      it 'returns empty string for failed transcriptions' do
+        audio_attachment # trigger creation
+
+        attachments = message.attachments
+        result = service.send(:extract_audio_transcriptions, attachments)
+        expect(result).to eq('')
+      end
+    end
+  end
+
+  describe 'private helper methods' do
+    describe '#text_part' do
+      it 'returns correct text part format' do
+        result = service.send(:text_part, 'Hello world')
+        expect(result).to eq({ type: 'text', text: 'Hello world' })
+      end
+    end
+
+    describe '#image_part' do
+      it 'returns correct image part format' do
+        result = service.send(:image_part, 'https://example.com/image.jpg')
+        expect(result).to eq({ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } })
+      end
+    end
+  end
+end