Files
chatwoot/spec/lib/captain/base_task_service_spec.rb
T
Aakash BakhleandGitHub 45f4b423ae fix: captain usage for BYOK OpenAI tasks (#14587)
# Pull Request Template

## Description

Fixes:
https://linear.app/chatwoot/issue/CW-7167/label-suggestions-bad-ux

## Type of change

Please delete options that are not relevant.

- [x] Bug fix (non-breaking change which fixes an issue)

## How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide
instructions so we can reproduce. Please also list any relevant details
for your test configuration.


## Checklist:

- [ ] My code follows the style guidelines of this project
- [ ] I have performed a self-review of my code
- [ ] I have commented on my code, particularly in hard-to-understand
areas
- [ ] I have made corresponding changes to the documentation
- [ ] My changes generate no new warnings
- [ ] I have added tests that prove my fix is effective or that my
feature works
- [ ] New and existing unit tests pass locally with my changes
- [ ] Any dependent changes have been merged and published in downstream
modules
2026-06-08 13:37:21 +05:30

394 lines
14 KiB
Ruby

require 'rails_helper'
RSpec.describe Captain::BaseTaskService do
let(:account) { create(:account) }
let(:inbox) { create(:inbox, account: account) }
let(:conversation) { create(:conversation, account: account, inbox: inbox) }
# Create a concrete test service class since BaseTaskService is abstract
let(:test_service_class) do
Class.new(described_class) do
def perform
{ message: 'Test response' }
end
def event_name
'test_event'
end
end
end
let(:service) { test_service_class.new(account: account, conversation_display_id: conversation.display_id) }
before do
create(:installation_config, name: 'CAPTAIN_OPEN_AI_API_KEY', value: 'test-key')
# Stub captain enabled check to allow OSS specs to test base functionality
# without enterprise module interference
allow(account).to receive(:feature_enabled?).and_call_original
allow(account).to receive(:feature_enabled?).with('captain_tasks').and_return(true)
allow(Integrations::Openai::KeyValidator).to receive(:valid?).and_return(true)
end
describe '#perform' do
it 'returns the expected result' do
result = service.perform
expect(result).to eq({ message: 'Test response' })
end
end
describe '#event_name' do
it 'raises NotImplementedError for base class' do
base_service = described_class.new(account: account, conversation_display_id: conversation.display_id)
expect { base_service.send(:event_name) }.to raise_error(NotImplementedError, /must implement #event_name/)
end
it 'returns custom event name in subclass' do
expect(service.send(:event_name)).to eq('test_event')
end
end
describe '#conversation' do
it 'finds conversation by display_id' do
expect(service.send(:conversation)).to eq(conversation)
end
it 'memoizes the conversation' do
expect(account.conversations).to receive(:find_by).once.and_return(conversation)
service.send(:conversation)
service.send(:conversation)
end
end
describe '#conversation_messages' do
let(:message1) { create(:message, conversation: conversation, message_type: :incoming, content: 'Hello', created_at: 1.hour.ago) }
let(:message2) { create(:message, conversation: conversation, message_type: :outgoing, content: 'Hi there', created_at: 30.minutes.ago) }
let(:message3) { create(:message, conversation: conversation, message_type: :incoming, content: 'How are you?', created_at: 10.minutes.ago) }
let(:private_message) { create(:message, conversation: conversation, message_type: :incoming, content: 'Private', private: true) }
before do
message1
message2
message3
private_message
end
it 'returns messages in array format with role and content' do
messages = service.send(:conversation_messages)
expect(messages).to be_an(Array)
expect(messages.length).to eq(3)
expect(messages[0]).to eq({ role: 'user', content: 'Hello' })
expect(messages[1]).to eq({ role: 'assistant', content: 'Hi there' })
expect(messages[2]).to eq({ role: 'user', content: 'How are you?' })
end
it 'excludes private messages' do
messages = service.send(:conversation_messages)
contents = messages.pluck(:content)
expect(contents).not_to include('Private')
end
it 'respects token limit' do
# Create messages that collectively exceed token limit
# Message validation max is 150000, so create multiple large messages
10.times do |i|
create(:message, conversation: conversation, message_type: :incoming,
content: 'a' * 100_000, created_at: i.minutes.ago)
end
messages = service.send(:conversation_messages)
total_length = messages.sum { |m| m[:content].length }
expect(total_length).to be <= Captain::BaseTaskService::TOKEN_LIMIT
end
it 'respects start_from offset for token counting' do
# With a start_from offset, fewer messages should fit
start_from = Captain::BaseTaskService::TOKEN_LIMIT - 100
messages = service.send(:conversation_messages, start_from: start_from)
total_length = messages.sum { |m| m[:content].length }
expect(total_length).to be <= 100
end
end
describe '#make_api_call' do
let(:model) { 'gpt-4' }
let(:messages) { [{ role: 'system', content: 'Test' }, { role: 'user', content: 'Hello' }] }
let(:mock_chat) { instance_double(RubyLLM::Chat) }
let(:mock_context) { instance_double(RubyLLM::Context, chat: mock_chat) }
let(:mock_response) { instance_double(RubyLLM::Message, content: 'Response', input_tokens: 10, output_tokens: 20) }
before do
allow(Llm::Config).to receive(:with_api_key).and_yield(mock_context)
allow(mock_chat).to receive(:with_instructions)
allow(mock_chat).to receive(:ask).and_return(mock_response)
end
context 'when captain_tasks is disabled' do
before do
allow(account).to receive(:feature_enabled?).with('captain_tasks').and_return(false)
end
it 'returns disabled error' do
result = service.send(:make_api_call, model: model, messages: messages)
expect(result[:error]).to eq(I18n.t('captain.disabled'))
expect(result[:error_code]).to eq(403)
end
it 'does not make API call' do
expect(Llm::Config).not_to receive(:with_api_key)
service.send(:make_api_call, model: model, messages: messages)
end
end
context 'when API key is not configured' do
before do
InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_API_KEY')&.destroy
# Clear memoized api_key
service.instance_variable_set(:@api_key, nil)
end
it 'returns api key missing error' do
result = service.send(:make_api_call, model: model, messages: messages)
expect(result[:error]).to eq(I18n.t('captain.api_key_missing'))
expect(result[:error_code]).to eq(401)
end
it 'does not make API call' do
expect(Llm::Config).not_to receive(:with_api_key)
service.send(:make_api_call, model: model, messages: messages)
end
end
it 'instruments the LLM call' do
expect(service).to receive(:instrument_llm_call).and_call_original
service.send(:make_api_call, model: model, messages: messages)
end
it 'returns formatted response with tokens' do
result = service.send(:make_api_call, model: model, messages: messages)
expect(result[:message]).to eq('Response')
expect(result[:usage]['prompt_tokens']).to eq(10)
expect(result[:usage]['completion_tokens']).to eq(20)
expect(result[:usage]['total_tokens']).to eq(30)
end
end
describe 'chat setup' do
let(:model) { 'gpt-4' }
let(:mock_chat) { instance_double(RubyLLM::Chat) }
let(:mock_context) { instance_double(RubyLLM::Context, chat: mock_chat) }
let(:mock_response) { instance_double(RubyLLM::Message, content: 'Response', input_tokens: 10, output_tokens: 20) }
before do
allow(Llm::Config).to receive(:with_api_key).and_yield(mock_context)
allow(mock_response).to receive(:input_tokens).and_return(10)
allow(mock_response).to receive(:output_tokens).and_return(20)
end
context 'with system instructions' do
let(:messages) { [{ role: 'system', content: 'You are helpful' }, { role: 'user', content: 'Hello' }] }
it 'applies system instructions to chat' do
expect(mock_chat).to receive(:with_instructions).with('You are helpful')
expect(mock_chat).to receive(:ask).with('Hello').and_return(mock_response)
service.send(:make_api_call, model: model, messages: messages)
end
end
context 'with conversation history' do
let(:messages) do
[
{ role: 'system', content: 'You are helpful' },
{ role: 'user', content: 'First message' },
{ role: 'assistant', content: 'First response' },
{ role: 'user', content: 'Second message' }
]
end
it 'adds conversation history before asking' do
expect(mock_chat).to receive(:with_instructions).with('You are helpful')
expect(mock_chat).to receive(:add_message).with(role: :user, content: 'First message').ordered
expect(mock_chat).to receive(:add_message).with(role: :assistant, content: 'First response').ordered
expect(mock_chat).to receive(:ask).with('Second message').and_return(mock_response)
service.send(:make_api_call, model: model, messages: messages)
end
end
context 'with single message' do
let(:messages) { [{ role: 'system', content: 'You are helpful' }, { role: 'user', content: 'Hello' }] }
it 'does not add conversation history' do
expect(mock_chat).to receive(:with_instructions).with('You are helpful')
expect(mock_chat).not_to receive(:add_message)
expect(mock_chat).to receive(:ask).with('Hello').and_return(mock_response)
service.send(:make_api_call, model: model, messages: messages)
end
end
end
describe 'error handling' do
let(:model) { 'gpt-4' }
let(:messages) { [{ role: 'user', content: 'Hello' }] }
let(:error) { StandardError.new('API Error') }
let(:exception_tracker) { instance_double(ChatwootExceptionTracker) }
before do
allow(Llm::Config).to receive(:with_api_key).and_raise(error)
allow(ChatwootExceptionTracker).to receive(:new).with(error, account: account).and_return(exception_tracker)
allow(exception_tracker).to receive(:capture_exception)
end
it 'tracks exceptions' do
expect(ChatwootExceptionTracker).to receive(:new).with(error, account: account).and_return(exception_tracker)
expect(exception_tracker).to receive(:capture_exception)
service.send(:make_api_call, model: model, messages: messages)
end
it 'returns error response' do
expect(exception_tracker).to receive(:capture_exception)
result = service.send(:make_api_call, model: model, messages: messages)
expect(result[:error]).to eq('API Error')
expect(result[:request_messages]).to eq(messages)
end
it 'tracks exceptions against the system key when an account hook exists' do
create(:integrations_hook, :openai, account: account, settings: { 'api_key' => 'hook-key' })
expect(Llm::Config).to receive(:with_api_key).with('test-key', api_base: anything).and_raise(error)
expect(ChatwootExceptionTracker).to receive(:new).with(error, account: account).and_return(exception_tracker)
expect(exception_tracker).to receive(:capture_exception)
result = service.send(:make_api_call, model: model, messages: messages)
expect(result[:error]).to eq('API Error')
expect(result[:request_messages]).to eq(messages)
end
end
describe '#api_key' do
context 'when openai hook is configured' do
let(:hook) { create(:integrations_hook, account: account, app_id: 'openai', status: 'enabled', settings: { 'api_key' => 'hook-key' }) }
before { hook }
it 'uses system api key by default' do
expect(service.send(:api_key)).to eq('test-key')
end
end
context 'when subclass opts into account OpenAI hook usage' do
let(:test_service_class) do
Class.new(described_class) do
def event_name
'test_event'
end
def use_account_openai_hook?
true
end
end
end
before do
create(:integrations_hook, account: account, app_id: 'openai', status: 'enabled', settings: { 'api_key' => 'hook-key' })
end
it 'uses api key from hook' do
expect(service.send(:api_key)).to eq('hook-key')
end
end
it 'uses account OpenAI hook for editor task services' do
create(:integrations_hook, account: account, app_id: 'openai', status: 'enabled', settings: { 'api_key' => 'hook-key' })
user = create(:user, account: account)
follow_up_context = {
'event_name' => 'professional',
'original_context' => 'Original text',
'last_response' => 'Last response'
}
editor_services = [
Captain::RewriteService.new(account: account, content: 'Text', operation: 'improve', conversation_display_id: conversation.display_id),
Captain::SummaryService.new(account: account, conversation_display_id: conversation.display_id),
Captain::ReplySuggestionService.new(account: account, conversation_display_id: conversation.display_id, user: user),
Captain::LabelSuggestionService.new(account: account, conversation_display_id: conversation.display_id),
Captain::FollowUpService.new(
account: account,
follow_up_context: follow_up_context,
user_message: 'Make it shorter',
conversation_display_id: conversation.display_id
)
]
editor_services.each do |editor_service|
expect(editor_service.send(:api_key)).to eq('hook-key')
end
end
context 'when openai hook is not configured' do
it 'uses system api key' do
expect(service.send(:api_key)).to eq('test-key')
end
end
context 'when no API key is configured' do
before do
InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_API_KEY')&.destroy
end
it 'returns nil' do
expect(service.send(:api_key)).to be_nil
end
end
end
describe '#prompt_from_file' do
it 'reads prompt from file' do
allow(Rails.root).to receive(:join).and_return(instance_double(Pathname, read: 'Test prompt content'))
expect(service.send(:prompt_from_file, 'test')).to eq('Test prompt content')
end
end
describe '#extract_original_context' do
it 'returns the most recent user message' do
messages = [
{ role: 'user', content: 'First question' },
{ role: 'assistant', content: 'First response' },
{ role: 'user', content: 'Follow-up question' }
]
result = service.send(:extract_original_context, messages)
expect(result).to eq('Follow-up question')
end
it 'returns nil when no user messages exist' do
messages = [
{ role: 'system', content: 'System prompt' },
{ role: 'assistant', content: 'Response' }
]
result = service.send(:extract_original_context, messages)
expect(result).to be_nil
end
it 'returns the only user message when there is just one' do
messages = [
{ role: 'system', content: 'System prompt' },
{ role: 'user', content: 'Single question' }
]
result = service.send(:extract_original_context, messages)
expect(result).to eq('Single question')
end
end
end