chat.conversation.send-message

Append a user turn to a chat conversation and return the assistant reply

End-to-end chat send-message pipeline.

Overview

Property	Value
Workflow type	Dag
Library	App-chat
Version	1.0

Input Schema

Field	Type	Required	Default	Description
conversation_uuid	uuid	Yes	—	ChatConversation.uuid (external surface)
user_uuid	uuid	No	—	User.uuid (external surface). Required for a member run; refused on an end-user run
user_message	string	Yes	—	Raw user message content (1-32000 chars)
provided_inputs	json	No	—	Optional resolved form payload for a prior pending_input
confirmed_action	json	No	—	Optional resolution of a prior pending_action
llm_connection_uuid	uuid	No	—	Optional LLM connection override; falls back to org default
stream	boolean	No	—	Streaming hint (metadata-only today)
deep_reasoning	boolean	No	—	Deep-reasoning mode: raise the dgi tool-loop budget so the model explores more of the catalog before answering
proactive	boolean	No	—	Proactive mode: after answering, surface 2-3 suggested next actions the user can take
workflow_run_uuid	uuid	No	—	Engine-stamped parent run UUID (DAG step target)

Output Schema

Field	Type	Required	Default	Description
conversation_uuid	uuid	No	—	—
user_uuid	uuid	No	—	—
user_message	string	No	—	—
provided_inputs	json	No	—	—
confirmed_action	json	No	—	—
llm_connection_uuid	uuid	No	—	—
stream	boolean	No	—	—
deep_reasoning	boolean	No	—	—
proactive	boolean	No	—	—
workflow_run_uuid	uuid	No	—	DAG run UUID; stamped by the engine's auto-generated 'start' transition.
validate	json	No	—	Result dict from the validate step
persist_user_message	json	No	—	Result dict from the persist-user-message step
dispatch_dgi	json	No	—	Result dict from the dispatch-dgi step
wait_and_persist_assistant	json	No	—	Result dict from the wait-and-persist-assistant step
return_payload	json	No	—	Result dict from the return-payload step
completed_at	string	No	—	—
failed_at	string	No	—	—
failure_reason	string	No	—	Engine-stamped human-readable failure reason
failed_at_state	string	No	—	Engine-stamped state when the workflow failed
failed_step	string	No	—	Engine-stamped step name (DAG path)
failed_layer	integer	No	—	Engine-stamped layer index (DAG path)
error	string	No	—	Engine-stamped exception message
error_type	string	No	—	Engine-stamped exception class name
comp_current_layer	integer	No	—	DAG framework: index of the layer currently being compensated
comp_queue	list	No	—	DAG framework: remaining layer indices queued for compensation
comp_retry_count	integer	No	—	DAG framework: retry count for the in-flight compensation step
compensation_trigger	string	No	—	DAG framework: 'failed'

DAG Layers

#	Layer	Steps	Compensation
1	validate	chat.conversation.send-message.validate	—
2	persist_user_message	chat.conversation.send-message.persist-user-message	—
3	dispatch_dgi	chat.conversation.send-message.dispatch-dgi	—
4	wait_and_persist_assistant	chat.conversation.send-message.wait-and-persist-assistant	—
5	return_payload	chat.conversation.send-message.return-payload	—

Execution Flow

Sub-workflows

Sub-workflow	Step name
chat.conversation.send-message.validate	validate
chat.conversation.send-message.persist-user-message	persist_user_message
chat.conversation.send-message.dispatch-dgi	dispatch_dgi
chat.conversation.send-message.wait-and-persist-assistant	wait_and_persist_assistant
chat.conversation.send-message.return-payload	return_payload

API Usage

POST /api/workflows/start
Content-Type: application/json

{
  "workflow_type": "chat.conversation.send-message",
  "initial_data": {
    "conversation_uuid": "value",
    "user_message": "value"
  }
}