Send Agent App Message

Send a user message to an Agent app. Agent apps use streaming responses only and may return message chunks, agent reasoning/tool events, generated files, and a final message_end event.

POST /chat-messages

Send Agent App Message

cURL

curl --request POST \
  --url https://{api_base_url}/chat-messages \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "inputs": {
    "customer_tier": "enterprise"
  },
  "query": "Summarize the attached product feedback and suggest next steps.",
  "response_mode": "streaming",
  "conversation_id": "",
  "user": "user-123",
  "files": [
    {
      "type": "document",
      "transfer_method": "local_file",
      "upload_file_id": "72fa9618-8f89-4a37-9b33-7e1178a24a67"
    }
  ]
}
'

200

streamingResponseAgent

"data: {\"event\": \"agent_thought\", \"id\": \"agent_thought_id_1\", \"task_id\": \"task123\", \"message_id\": \"msg123\", \"conversation_id\": \"conv123\", \"position\": 1, \"thought\": \"Thinking about calling a tool...\", \"tool\": \"dalle3\", \"tool_input\": \"{\\\"dalle3\\\": {\\\"prompt\\\": \\\"a cute cat\\\"}}\", \"created_at\": 1705395332} data: {\"event\": \"message_file\", \"task_id\": \"task123\", \"message_id\": \"msg123\", \"conversation_id\": \"conv123\", \"id\": \"file_id_1\", \"type\": \"image\", \"belongs_to\": \"assistant\", \"url\": \"https://example.com/cat.png\", \"created_at\": 1705395332} data: {\"event\": \"agent_message\", \"task_id\": \"task123\", \"message_id\": \"msg123\", \"conversation_id\": \"conv123\", \"answer\": \"Here is the image: \", \"created_at\": 1705395333} data: {\"event\": \"message_end\", \"task_id\":\"task123\", \"message_id\": \"msg123\", \"conversation_id\": \"conv123\", \"metadata\": {\"usage\": {\"total_tokens\": 50, \"latency\": 2.5}}}"

Authorizations

Authorization · string · header · required

API Key authentication. For all API requests, include your API Key in the Authorization HTTP Header, prefixed with Bearer. Example: Authorization: Bearer {API_KEY}. Strongly recommend storing your API Key on the server-side, not shared or stored on the client-side, to avoid possible API-Key leakage that can lead to serious consequences.

Body

application/json

Request body to send an Agent app message.

Request body for sending a message to an Agent app.

query · string · required

User message or task for the Agent app.

inputs · object · required

Values for variables configured in the Agent Soul. Call the Get Agent App Parameters API to discover expected variable names and input controls.

user · string · required

End-user identifier, unique within this Agent app. Conversations, messages, and uploaded files are scoped to the same user value.

response_mode · enum<string> · default: streaming

Agent apps support streaming mode only. Set this field to streaming or omit it.

Available options: streaming

conversation_id · string

Conversation ID to continue. Omit this field or pass an empty string to start a new conversation, then reuse the returned conversation_id in later requests.

files · object[]

Files for multimodal Agent input. To attach a local file, upload it with the Upload File API first, then pass the returned file id as upload_file_id with transfer_method: local_file.

Show child attributes

auto_generate_name · boolean · default: true

Auto-generate conversation title. If false, use the Rename Conversation API with auto_generate: true for async title generation.

Response

200

text/event-stream

Successful streaming response. Agent apps return text/event-stream Server-Sent Events. Blocking JSON responses are not supported for Agent apps.

A stream of Server-Sent Events (SSE).

Parsing: Each event is a line prefixed with data: followed by a JSON object, terminated by \n\n. Strip the data: prefix before parsing the JSON, then read the event field to determine the event type. Ignore ping events, which arrive as event: ping lines (no data: payload) every 10 seconds to keep the connection alive.

Stream lifecycle: The reply streams as message events (Chatbot apps), or as agent_thought and agent_message events (Agent apps), and ends with message_end. When text-to-speech auto-play is enabled, tts_message events interleave and tts_message_end becomes the final event.

Events: Apart from ping, every event includes conversation_id, message_id, and created_at (Unix epoch seconds); all but error also include task_id.

Reply events

Event	App	Fires on	Key fields
message	Chatbot	each answer chunk (concatenate in order)	answer
agent_message	Agent	each answer chunk (concatenate in order)	answer
agent_thought	Agent	each reasoning or tool-call step	position, thought, tool, tool_input (JSON), observation, message_files
message_replace	Chatbot, Agent	output moderation replaces the answer so far	answer
message_file	Chatbot, Agent	the assistant returns a file	type, belongs_to, url
message_end	Chatbot, Agent	the answer is complete	metadata (usage, retriever_resources)
tts_message, tts_message_end	Chatbot, Agent	audio chunk / end, when TTS auto-play is on	audio

Transport events

Event	App	Fires on	Key fields
error	Chatbot, Agent	a failure ends the stream; HTTP stays 200	status (e.g. 400), code (e.g. invalid_param), message
ping	Chatbot, Agent	keep-alive every 10 seconds	none