Records a `Captain::Session` row for every Captain V2 assistant response delivered in a conversation, so we can show how a response was generated and report on credit, FAQ, and document usage. Stacked on #14970 (the `captain_sessions` model). ## What changed - `FaqLookupTool` now records the retrieved FAQ ids (and their backing document ids) into the shared run state, accumulated across tool calls. - `AgentRunnerService` exposes the raw ai-agents run result via `last_run_result`; the `generate_response` return shape is unchanged, so the playground path is unaffected. - New `Captain::Assistant::SessionCaptureService` builds the session: scenario resolved from the answering agent name, model from `assistant.agent_model`, token usage plus the trimmed current-turn conversation history stored in `run_context`. - `ResponseBuilderJob` captures after delivery: `credits_consumed` mirrors the actual charge (1.0 for a billed response, 0.0 for handoffs, where the session points at the customer-facing handoff message). Capture runs outside the delivery transaction and swallows its own failures, so a logging bug can never block or roll back a customer reply. V1 responses and copilot are out of scope; copilot capture comes next. ## How to test On an account with `captain_integration_v2` enabled and an inbox connected to an assistant with approved FAQs, send a customer message on a pending conversation. After the assistant replies, a `Captain::Session` row should exist with the conversation as subject, the reply message as result, the FAQs/documents used, and the run context for that turn. Asking for a human agent should produce a zero-credit session pointing at the handoff message. <img width="2428" height="1058" alt="CleanShot 2026-07-15 at 17 25 40@2x" src="https://github.com/user-attachments/assets/d8e44923-c17b-494f-8c33-c8fa4219438c" />
196 lines
7.0 KiB
Ruby
196 lines
7.0 KiB
Ruby
require 'agents'
|
|
require 'agents/instrumentation'
|
|
|
|
class Captain::Assistant::AgentRunnerService
|
|
include Integrations::LlmInstrumentationConstants
|
|
include Captain::Assistant::RunnerCallbacksHelper
|
|
include Captain::Assistant::TracePayloadHelper
|
|
include Captain::Assistant::RunnerStateHelper
|
|
|
|
attr_reader :last_run_result
|
|
|
|
def initialize(assistant:, conversation: nil, callbacks: {}, source: nil)
|
|
@assistant = assistant
|
|
@conversation = conversation
|
|
@callbacks = callbacks
|
|
@source = source
|
|
@handoff_tool_called = false
|
|
end
|
|
|
|
def generate_response(message_history: [])
|
|
message_to_process, context = run_payload(message_history)
|
|
@last_run_result = runner.run(message_to_process, context: context, max_turns: 10)
|
|
|
|
process_agent_result(@last_run_result)
|
|
rescue StandardError => e
|
|
# In rake/local runs, conversation may not be present, so account is optional here.
|
|
ChatwootExceptionTracker.new(e, account: @conversation&.account).capture_exception
|
|
Rails.logger.error "[Captain V2] AgentRunnerService error: #{e.message}"
|
|
Rails.logger.error e.backtrace.join("\n")
|
|
|
|
error_response(e.message)
|
|
end
|
|
|
|
private
|
|
|
|
def build_context(message_history)
|
|
conversation_history = message_history.map do |msg|
|
|
content = msg[:content]
|
|
# Preserve multimodal arrays (with image_url entries) as-is for the runner to restore with attachments.
|
|
# Only extract text from non-array formats (hashes from agent structured output, plain strings).
|
|
content = extract_text_from_content(content) unless content.is_a?(Array)
|
|
|
|
{
|
|
role: msg[:role].to_sym,
|
|
content: content,
|
|
agent_name: msg[:agent_name]
|
|
}
|
|
end
|
|
|
|
{
|
|
session_id: "#{@assistant.account_id}_#{@conversation&.display_id}",
|
|
conversation_history: conversation_history,
|
|
state: build_state
|
|
}
|
|
end
|
|
|
|
def extract_last_user_message(message_history)
|
|
last_user_msg = message_history.reverse.find { |msg| msg[:role] == 'user' }
|
|
return '' if last_user_msg.blank?
|
|
|
|
content = last_user_msg[:content]
|
|
return extract_text_from_content(content) unless content.is_a?(Array)
|
|
|
|
text, attachments = Captain::OpenAiMessageBuilderService.extract_text_and_attachments(content)
|
|
return text if attachments.blank?
|
|
|
|
RubyLLM::Content.new(text, attachments)
|
|
end
|
|
|
|
def message_history_without_last_user_message(message_history)
|
|
last_user_index = message_history.rindex { |msg| msg[:role] == 'user' }
|
|
return message_history if last_user_index.nil?
|
|
|
|
message_history.reject.with_index { |_msg, index| index == last_user_index }
|
|
end
|
|
|
|
def extract_text_from_content(content)
|
|
# Handle structured output from agents
|
|
return content[:response] || content['response'] || content.to_s if content.is_a?(Hash)
|
|
|
|
return content unless content.is_a?(Array)
|
|
|
|
text_parts = content.select { |part| part[:type] == 'text' }.pluck(:text)
|
|
text_parts.join(' ')
|
|
end
|
|
|
|
def process_agent_result(result)
|
|
Rails.logger.info "[Captain V2] Agent result: #{result.inspect}"
|
|
output = result.output
|
|
response = output.is_a?(Hash) ? output.with_indifferent_access : { 'response' => output.to_s, 'reasoning' => 'Processed by agent' }
|
|
response['agent_name'] = result.context&.dig(:current_agent)
|
|
response['handoff_tool_called'] = result.context&.dig(:captain_v2_handoff_tool_called) || false
|
|
response
|
|
end
|
|
|
|
def error_response(error_message)
|
|
{
|
|
'response' => 'conversation_handoff',
|
|
'reasoning' => "Error occurred: #{error_message}",
|
|
'handoff_tool_called' => @handoff_tool_called
|
|
}
|
|
end
|
|
|
|
def build_and_wire_agents
|
|
assistant_agent = @assistant.agent
|
|
scenario_agents = @assistant.scenarios.enabled.map(&:agent)
|
|
|
|
assistant_agent.register_handoffs(*scenario_agents) if scenario_agents.any?
|
|
scenario_agents.each { |scenario_agent| scenario_agent.register_handoffs(assistant_agent) }
|
|
|
|
[assistant_agent] + scenario_agents
|
|
end
|
|
|
|
def install_instrumentation(runner)
|
|
return unless ChatwootApp.otel_enabled?
|
|
|
|
Agents::Instrumentation.install(
|
|
runner,
|
|
tracer: OpentelemetryConfig.tracer,
|
|
trace_name: 'llm.captain_v2',
|
|
span_attributes: {
|
|
ATTR_LANGFUSE_TAGS => ['captain_v2'].to_json
|
|
},
|
|
attribute_provider: Captain::Assistant::InstrumentationAttributeProvider.new(self)
|
|
)
|
|
register_trace_input_callback(runner)
|
|
end
|
|
|
|
def dynamic_trace_attributes(context_wrapper)
|
|
state = context_wrapper&.context&.dig(:state) || {}
|
|
conversation = state[:conversation] || {}
|
|
trace_input = context_wrapper&.context&.dig(:captain_v2_trace_input)
|
|
|
|
{
|
|
ATTR_LANGFUSE_USER_ID => state[:account_id],
|
|
format(ATTR_LANGFUSE_METADATA, 'assistant_id') => state[:assistant_id],
|
|
format(ATTR_LANGFUSE_METADATA, 'conversation_display_id') => conversation[:display_id],
|
|
format(ATTR_LANGFUSE_METADATA, 'channel_type') => state[:channel_type],
|
|
format(ATTR_LANGFUSE_METADATA, 'source') => state[:source],
|
|
ATTR_LANGFUSE_TRACE_INPUT => trace_input,
|
|
ATTR_LANGFUSE_OBSERVATION_INPUT => trace_input
|
|
}.compact.transform_values(&:to_s)
|
|
end
|
|
|
|
def add_usage_metadata_callback(runner)
|
|
handoff_tool_name = Captain::Tools::HandoffTool.new(@assistant).name
|
|
|
|
# Tool tracking always runs — process_response in the job consumes the resulting
|
|
# handoff_tool_called flag regardless of whether OTEL is enabled.
|
|
runner.on_tool_complete do |tool_name, _tool_result, context_wrapper|
|
|
track_handoff_usage(tool_name, handoff_tool_name, context_wrapper)
|
|
end
|
|
|
|
if ChatwootApp.otel_enabled?
|
|
runner.on_run_complete do |_agent_name, _result, context_wrapper|
|
|
write_credits_used_metadata(context_wrapper)
|
|
end
|
|
end
|
|
runner
|
|
end
|
|
|
|
def track_handoff_usage(tool_name, handoff_tool_name, context_wrapper)
|
|
return unless context_wrapper&.context
|
|
return unless tool_name.to_s == handoff_tool_name
|
|
|
|
# Mirror the flag onto the instance so error_response can surface it even when
|
|
# the runner raises before returning a result (the context is unreachable then).
|
|
context_wrapper.context[:captain_v2_handoff_tool_called] = true
|
|
@handoff_tool_called = true
|
|
end
|
|
|
|
def write_credits_used_metadata(context_wrapper)
|
|
root_span = context_wrapper&.context&.dig(:__otel_tracing, :root_span)
|
|
return unless root_span
|
|
|
|
root_span.set_attribute(format(ATTR_LANGFUSE_METADATA, 'credit_used'), @handoff_tool_called ? 'false' : 'true')
|
|
end
|
|
|
|
def runner
|
|
@runner ||= begin
|
|
configured_runner = Agents::Runner.with_agents(*build_and_wire_agents)
|
|
configured_runner = add_usage_metadata_callback(configured_runner)
|
|
configured_runner = add_callbacks_to_runner(configured_runner) if @callbacks.any?
|
|
install_instrumentation(configured_runner)
|
|
configured_runner
|
|
end
|
|
end
|
|
|
|
def run_payload(message_history)
|
|
message_to_process = extract_last_user_message(message_history)
|
|
context = build_context(message_history_without_last_user_message(message_history))
|
|
enrich_context_with_trace_payload!(context, message_history, message_to_process)
|
|
[message_to_process, context]
|
|
end
|
|
end
|