LeadMagnet/leadchat
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

feat: support multimodal user messages in captain v2 (#13581)

Browse Source 
Extract and pass image attachments from the latest user message to the
runner,
excluding the last user message from the context for processing.

Fixes #13588 

# Pull Request Template

## Description

Adds image support to captain v2

## 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.

specs and local testing

<img width="754" height="1008" alt="image"
src="https://github.com/user-attachments/assets/914cbc2c-9d30-42d0-87d4-9e5430845c87"
/>

langfuse also shows media correctly with the instrumentation code:
<img width="1800" height="1260" alt="image"
src="https://github.com/user-attachments/assets/ce0f5fa6-b1a5-42ec-a213-9a82b1751037"
/>


## 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

---------

Co-authored-by: Shivam Mishra <scm.mymail@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Aakash Bakhle
2026-02-24 19:37:41 +05:30
committed by GitHub
parent 6be95e79f8
commit 7cec4ebaae
6 changed files with 264 additions and 62 deletions
Download Patch File Download Diff File Expand all files Collapse all files

131
spec/enterprise/services/captain/assistant/agent_runner_service_spec.rb
View File

@@ -74,12 +74,11 @@ RSpec.describe Captain::Assistant::AgentRunnerService do
end
it 'runs agent with extracted user message and context' do
expected_context = {
expected_context = hash_including(
session_id: "#{account.id}_#{conversation.display_id}",
conversation_history: [
{ role: :user, content: 'Hello there', agent_name: nil },
{ role: :assistant, content: 'Hi! How can I help you?', agent_name: 'Assistant' },
{ role: :user, content: 'I need help with my account', agent_name: nil }
{ role: :assistant, content: 'Hi! How can I help you?', agent_name: 'Assistant' }
],
state: hash_including(
account_id: account.id,
@@ -87,7 +86,7 @@ RSpec.describe Captain::Assistant::AgentRunnerService do
conversation: hash_including(id: conversation.id),
contact: hash_including(id: contact.id)
)
}
)
expect(mock_runner).to receive(:run).with(
'I need help with my account',
@@ -98,6 +97,71 @@ RSpec.describe Captain::Assistant::AgentRunnerService do
service.generate_response(message_history: message_history)
end
context 'when the latest user message is multimodal' do
let(:multimodal_message_history) do
[
{ role: 'assistant', content: 'Please share a screenshot' },
{
role: 'user',
content: [
{ type: 'text', text: 'What does this error mean?' },
{ type: 'image_url', image_url: { url: 'https://example.com/error.png' } }
]
}
]
end
it 'passes image attachments to the runner input' do
expect(mock_runner).to receive(:run) do |input, context:, max_turns:|
expect(input).to be_a(RubyLLM::Content)
expect(input.text).to eq('What does this error mean?')
expect(input.attachments.first.source.to_s).to eq('https://example.com/error.png')
expect(context[:conversation_history]).to eq([{ role: :assistant, content: 'Please share a screenshot', agent_name: nil }])
expect(max_turns).to eq(100)
end
service.generate_response(message_history: multimodal_message_history)
end
it 'preserves multimodal content in earlier history messages' do
history_with_prior_image = [
{
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' }
]
expect(mock_runner).to receive(:run) do |input, context:, max_turns:|
expect(input).to eq('It still does not work')
# The earlier user message with the image should preserve the multimodal array
first_history_msg = context[:conversation_history].first
expect(first_history_msg[:content]).to be_a(Array)
expect(first_history_msg[:content]).to include(
{ type: 'text', text: 'Here is my error screenshot' },
{ type: 'image_url', image_url: { url: 'https://example.com/error.png' } }
)
expect(max_turns).to eq(100)
end
service.generate_response(message_history: history_with_prior_image)
end
it 'stores multimodal trace payloads in runner context' do
expect(mock_runner).to receive(:run) do |_input, context:, max_turns:|
expect(context[:captain_v2_trace_input]).to include('image_url')
expect(context[:captain_v2_trace_current_input]).to include('image_url')
expect(max_turns).to eq(100)
end
service.generate_response(message_history: multimodal_message_history)
end
end
it 'processes and formats agent result' do
result = service.generate_response(message_history: message_history)
@@ -197,22 +261,21 @@ RSpec.describe Captain::Assistant::AgentRunnerService do
end
context 'with multimodal content' do
let(:multimodal_message_history) do
let(:multimodal_content) do
[
{
role: 'user',
content: [
{ type: 'text', text: 'Can you help with this image?' },
{ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } }
]
}
{ type: 'text', text: 'Can you help with this image?' },
{ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } }
]
end
it 'extracts text content from multimodal messages' do
let(:multimodal_message_history) do
[{ role: 'user', content: multimodal_content }]
end
it 'preserves multimodal arrays in conversation history for image context retention' do
context = service.send(:build_context, multimodal_message_history)
expect(context[:conversation_history].first[:content]).to eq('Can you help with this image?')
expect(context[:conversation_history].first[:content]).to eq(multimodal_content)
end
end
end
@@ -225,6 +288,24 @@ RSpec.describe Captain::Assistant::AgentRunnerService do
expect(result).to eq('I need help with my account')
end
it 'returns multimodal content with image attachments for the runner input' do
multimodal_message_history = [
{
role: 'user',
content: [
{ type: 'text', text: 'Can you check this screenshot?' },
{ type: 'image_url', image_url: { url: 'https://example.com/image.jpg' } }
]
}
]
result = service.send(:extract_last_user_message, multimodal_message_history)
expect(result).to be_a(RubyLLM::Content)
expect(result.text).to eq('Can you check this screenshot?')
expect(result.attachments.first.source.to_s).to eq('https://example.com/image.jpg')
end
end
describe '#extract_text_from_content' do
@@ -256,6 +337,28 @@ RSpec.describe Captain::Assistant::AgentRunnerService do
end
end
describe '#dynamic_trace_attributes' do
subject(:service) { described_class.new(assistant: assistant, conversation: conversation) }
it 'adds serialized trace input attributes when present in context' do
context = {
state: {
account_id: account.id,
assistant_id: assistant.id,
conversation: { id: conversation.id, display_id: conversation.display_id }
},
captain_v2_trace_input: '[{"role":"user","content":[{"type":"image_url","image_url":{"url":"https://example.com/image.jpg"}}]}]'
}
context_wrapper = Struct.new(:context).new(context)
attributes = service.send(:dynamic_trace_attributes, context_wrapper)
expect(attributes['langfuse.trace.input']).to include('image_url')
expect(attributes['langfuse.observation.input']).to include('image_url')
expect(attributes['langfuse.user.id']).to eq(account.id.to_s)
end
end
describe '#build_state' do
subject(:service) { described_class.new(assistant: assistant, conversation: conversation) }