feat(ee): Add transcription support for audio messages (#11670)

<img width="419" alt="Screenshot 2025-06-03 at 4 25 37 PM"
src="https://github.com/user-attachments/assets/4b6ddd11-9b91-4981-a571-83746cc4d40b"
/>


Fixes https://github.com/chatwoot/chatwoot/issues/10182

---------

Co-authored-by: Sojan Jose <sojan@pepalo.com>

spec/enterprise/jobs/messages/audio_transcription_job_spec.rb

@@ -0,0 +1,41 @@
+require 'rails_helper'
+
+RSpec.describe Messages::AudioTranscriptionJob do
+  subject(:job) { described_class.perform_later(attachment_id) }
+
+  let(:message) { create(:message) }
+  let(:attachment) do
+    message.attachments.create!(
+      account_id: message.account_id,
+      file_type: :audio,
+      file: fixture_file_upload('public/audio/widget/ding.mp3')
+    )
+  end
+  let(:attachment_id) { attachment.id }
+  let(:conversation) { message.conversation }
+  let(:transcription_service) { instance_double(Messages::AudioTranscriptionService) }
+
+  it 'enqueues the job' do
+    expect { job }.to have_enqueued_job(described_class)
+      .with(attachment_id)
+      .on_queue('low')
+  end
+
+  context 'when performing the job' do
+    before do
+      allow(Messages::AudioTranscriptionService).to receive(:new).with(attachment).and_return(transcription_service)
+      allow(transcription_service).to receive(:perform)
+    end
+
+    it 'calls AudioTranscriptionService with the attachment' do
+      expect(Messages::AudioTranscriptionService).to receive(:new).with(attachment)
+      expect(transcription_service).to receive(:perform)
+      described_class.perform_now(attachment_id)
+    end
+
+    it 'does nothing when attachment is not found' do
+      expect(Messages::AudioTranscriptionService).not_to receive(:new)
+      described_class.perform_now(999_999)
+    end
+  end
+end

spec/enterprise/services/messages/audio_transcription_service_spec.rb

@@ -0,0 +1,70 @@
+require 'rails_helper'
+
+RSpec.describe Messages::AudioTranscriptionService, type: :service do
+  let(:account) { create(:account, audio_transcriptions: true) }
+  let(:conversation) { create(:conversation, account: account) }
+  let(:message) { create(:message, conversation: conversation) }
+  let(:attachment) { message.attachments.create!(account: account, file_type: :audio) }
+
+  before do
+    # Create required installation configs
+    create(:installation_config, name: 'CAPTAIN_OPEN_AI_API_KEY', value: 'test-api-key')
+    create(:installation_config, name: 'CAPTAIN_OPEN_AI_MODEL', value: 'gpt-4o-mini')
+
+    # Mock usage limits for transcription to be available
+    allow(account).to receive(:usage_limits).and_return({ captain: { responses: { current_available: 100 } } })
+  end
+
+  describe '#perform' do
+    let(:service) { described_class.new(attachment) }
+
+    context 'when transcription is successful' do
+      before do
+        # Mock can_transcribe? to return true and transcribe_audio method
+        allow(service).to receive(:can_transcribe?).and_return(true)
+        allow(service).to receive(:transcribe_audio).and_return('Hello world transcription')
+      end
+
+      it 'returns successful transcription' do
+        result = service.perform
+        expect(result).to eq({ success: true, transcriptions: 'Hello world transcription' })
+      end
+    end
+
+    context 'when audio transcriptions are disabled' do
+      before do
+        account.update!(audio_transcriptions: false)
+      end
+
+      it 'returns error for transcription limit exceeded' do
+        result = service.perform
+        expect(result).to eq({ error: 'Transcription limit exceeded' })
+      end
+    end
+
+    context 'when attachment already has transcribed text' do
+      before do
+        attachment.update!(meta: { transcribed_text: 'Existing transcription' })
+        allow(service).to receive(:can_transcribe?).and_return(true)
+      end
+
+      it 'returns existing transcription without calling API' do
+        result = service.perform
+        expect(result).to eq({ success: true, transcriptions: 'Existing transcription' })
+      end
+    end
+
+    context 'when transcription fails' do
+      before do
+        allow(service).to receive(:can_transcribe?).and_return(true)
+        allow(service).to receive(:transcribe_audio).and_raise(StandardError.new('API error'))
+        allow(ChatwootExceptionTracker).to receive(:new).and_return(instance_double(ChatwootExceptionTracker, capture_exception: nil))
+      end
+
+      it 'returns error response' do
+        result = service.perform
+        expect(result).to eq({ error: 'Transcription failed: API error' })
+      end
+    end
+  end
+end