chatwoot/spec/enterprise/services/captain/llm/assistant_chat_service_spec.rb
Aakash Bakhle eaffad12e7
feat(langfuse): propagate observation metadata for evals (#14634)
# Pull Request Template

## Description

We need to pass on trace level attributes down to the spans inside them
like tool calls, observations, etc.
This way, we can filter observations based on trace level attributes.


## Type of change

- [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.

Attributes added to observation metadata for easy filtering
<img width="1327" height="708" alt="image"
src="https://github.com/user-attachments/assets/8f1d1bf8-cde4-481d-a2c2-7920ad2fc52e"
/>

added a `generation_stage` to differentiate llm_calls that call tools vs
those that generate a `final_response`
<img width="1806" height="968" alt="CleanShot 2026-06-03 at 15 11 09@2x"
src="https://github.com/user-attachments/assets/db1fa8e0-7f2d-404b-a719-27a16d400442"
/>


propagated attributes to tool calls for future use
<img width="903" height="517" alt="image"
src="https://github.com/user-attachments/assets/edc61ce8-93db-465c-a66e-043138e2dc15"
/>



## Checklist:

- [x] My code follows the style guidelines of this project
- [x] I have performed a self-review of my code
- [x] I have commented on my code, particularly in hard-to-understand
areas
- [ ] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
- [x] I have added tests that prove my fix is effective or that my
feature works
- [x] New and existing unit tests pass locally with my changes
- [x] Any dependent changes have been merged and published in downstream
modules
2026-06-03 16:45:19 +05:30

241 lines
9.2 KiB
Ruby

require 'rails_helper'
RSpec.describe Captain::Llm::AssistantChatService do
let(:account) { create(:account) }
let(:assistant) { create(:captain_assistant, account: account) }
let(:conversation) { create(:conversation, account: account) }
let(:mock_chat) { instance_double(RubyLLM::Chat) }
let(:mock_response) do
instance_double(
RubyLLM::Message,
content: '{"response": "I can see the image shows a pricing table", "reasoning": "Analyzed the image"}'
)
end
before do
create(:installation_config, name: 'CAPTAIN_OPEN_AI_API_KEY', value: 'test-key')
allow(RubyLLM).to receive(:chat).and_return(mock_chat)
allow(mock_chat).to receive(:with_temperature).and_return(mock_chat)
allow(mock_chat).to receive(:with_params).and_return(mock_chat)
allow(mock_chat).to receive(:with_tool).and_return(mock_chat)
allow(mock_chat).to receive(:with_instructions).and_return(mock_chat)
allow(mock_chat).to receive(:add_message).and_return(mock_chat)
allow(mock_chat).to receive(:on_end_message).and_return(mock_chat)
allow(mock_chat).to receive(:on_tool_call).and_return(mock_chat)
allow(mock_chat).to receive(:on_tool_result).and_return(mock_chat)
allow(mock_chat).to receive(:messages).and_return([])
end
describe 'instrumentation metadata' do
it 'passes channel_type to the agent session instrumentation' do
service = described_class.new(assistant: assistant, conversation: conversation)
expect(service).to receive(:instrument_agent_session).with(
hash_including(metadata: hash_including(channel_type: conversation.inbox.channel_type))
).and_yield
allow(mock_chat).to receive(:ask).and_return(mock_response)
service.generate_response(message_history: [{ role: 'user', content: 'Hello' }])
end
it 'marks final response generations for observation-level evaluators' do
service = described_class.new(assistant: assistant, conversation: conversation)
message = instance_double(RubyLLM::Message, content: 'Final answer', input_tokens: 10, output_tokens: 20, tool_calls: {})
attributes = service.send(:generation_attributes, mock_chat, message)
expect(attributes['langfuse.observation.metadata.generation_stage']).to eq('final_response')
end
it 'marks tool call generations separately from final responses' do
service = described_class.new(assistant: assistant, conversation: conversation)
message = instance_double(
RubyLLM::Message,
content: '',
input_tokens: 10,
output_tokens: 20,
tool_calls: { 'call_1' => instance_double(RubyLLM::ToolCall) }
)
attributes = service.send(:generation_attributes, mock_chat, message)
expect(attributes['langfuse.observation.metadata.generation_stage']).to eq('tool_call')
end
end
describe 'image analysis' do
context 'when user sends a message with an image attachment' do
let(:message_history) do
[
{
role: 'user',
content: [
{ type: 'text', text: 'What do you see in this image?' },
{ type: 'image_url', image_url: { url: 'https://example.com/screenshot.png' } }
]
}
]
end
it 'sends the image to the LLM for analysis' do
expect(mock_chat).to receive(:ask).with(
'What do you see in this image?',
with: ['https://example.com/screenshot.png']
).and_return(mock_response)
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: message_history)
end
end
context 'when user sends only an image without text' do
let(:message_history) do
[
{
role: 'user',
content: [
{ type: 'image_url', image_url: { url: 'https://example.com/photo.jpg' } }
]
}
]
end
it 'sends the image to the LLM with nil text' do
expect(mock_chat).to receive(:ask).with(
nil,
with: ['https://example.com/photo.jpg']
).and_return(mock_response)
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: message_history)
end
end
context 'when user sends a plain text message' do
let(:message_history) do
[
{ role: 'user', content: 'Hello, how can you help me?' }
]
end
it 'sends the text without attachments' do
expect(mock_chat).to receive(:ask).with('Hello, how can you help me?').and_return(mock_response)
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: message_history)
end
end
end
describe 'conversation history with images' do
context 'when previous messages contain images' do
let(:message_history) do
[
{
role: 'user',
content: [
{ type: 'text', text: 'Here is my error screenshot' },
{ type: 'image_url', image_url: { url: 'https://example.com/error.png' } }
]
},
{ role: 'assistant', content: 'I see the error. Try restarting.' },
{ role: 'user', content: 'It still does not work' }
]
end
it 'includes images from conversation history in context' do
# First historical message should include the image via RubyLLM::Content
expect(mock_chat).to receive(:add_message) do |args|
expect(args[:role]).to eq(:user)
expect(args[:content]).to be_a(RubyLLM::Content)
expect(args[:content].text).to eq('Here is my error screenshot')
expect(args[:content].attachments.first.source.to_s).to eq('https://example.com/error.png')
end.ordered
# Second historical message is plain text
expect(mock_chat).to receive(:add_message).with(
role: :assistant,
content: 'I see the error. Try restarting.'
).ordered
# Current message asked via chat.ask
expect(mock_chat).to receive(:ask).with('It still does not work').and_return(mock_response)
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: message_history)
end
end
end
describe 'contact attributes in system prompt' do
let(:contact) { create(:contact, account: account, name: 'Diep Bui', email: 'diep@example.com', custom_attributes: { 'plan' => 'pro' }) }
let(:conversation) { create(:conversation, account: account, contact: contact) }
context 'when feature_contact_attributes is enabled' do
before { assistant.update!(config: assistant.config.merge('feature_contact_attributes' => true)) }
it 'includes contact information in the system prompt' do
allow(mock_chat).to receive(:ask).and_return(mock_response)
expect(mock_chat).to receive(:with_instructions).with(a_string_including('[Contact Information]')) do |_instructions|
mock_chat
end
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: [{ role: 'user', content: 'Hello' }])
end
it 'includes custom attributes in the system prompt' do
allow(mock_chat).to receive(:ask).and_return(mock_response)
expect(mock_chat).to receive(:with_instructions).with(a_string_including('plan: pro')) do |_instructions|
mock_chat
end
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: [{ role: 'user', content: 'Hello' }])
end
end
context 'when feature_contact_attributes is disabled' do
it 'does not include contact information in the system prompt' do
allow(mock_chat).to receive(:ask).and_return(mock_response)
expect(mock_chat).to receive(:with_instructions).with(satisfy { |s| s.exclude?('[Contact Information]') }) do |_instructions|
mock_chat
end
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: [{ role: 'user', content: 'Hello' }])
end
end
end
describe 'account custom instructions in system prompt' do
before do
assistant.update!(config: assistant.config.merge('instructions' => 'if user enters 1112234 suggest handoff'))
end
it 'adds custom instructions in a separate delimited section' do
allow(mock_chat).to receive(:ask).and_return(mock_response)
expect(mock_chat).to receive(:with_instructions).with(
a_string_including(
'<account_custom_instructions>',
'if user enters 1112234 suggest handoff',
'</account_custom_instructions>'
)
) do |instructions|
expect(instructions).not_to include('<custom-instructions>')
expect(instructions.index('<account_custom_instructions>')).to be < instructions.index('```json')
mock_chat
end
service = described_class.new(assistant: assistant, conversation: conversation)
service.generate_response(message_history: [{ role: 'user', content: 'Hello' }])
end
end
end