Poll Status

Multimodal generation such as image and video generation usually takes time. The create-task endpoint only creates an asynchronous task; final status and results are obtained by polling.

Request

Status queries can use the status_url from the create-task response, or build the request path with BASE_URL, endpoint, and request_id.

Use status_url

curl -sS "${STATUS_URL}" \
  "${COMMON_HEADERS[@]}"

Query status by request_id

curl -sS "${BASE_URL}/v1/queue/${ENDPOINT}/requests/${REQUEST_ID}/status" \
  "${COMMON_HEADERS[@]}"

Response

The status query response indicates whether the task still needs to wait and provides the basis for retrieving results or handling errors.

status	Description	Recommended next step
IN_QUEUE	The task has been submitted and is waiting to run	Continue polling
IN_PROGRESS	The task is running	Continue polling
COMPLETED	The task has reached a terminal state	Check whether error exists; if not, retrieve the result

Tip

Recommended polling interval is every 2-5 seconds. For long videos or slower models, use 5-10 seconds.

IN_QUEUE response

202

{
  "status": "IN_QUEUE",
  "request_id": "mmsu_01hxyz...",
  "queue_position": 0,
  "response_url": "https://.../response"
}

IN_PROGRESS response

202

{
  "status": "IN_PROGRESS",
  "request_id": "mmsu_01hxyz...",
  "response_url": "https://.../response"
}

COMPLETED response

COMPLETED only means the task has reached a terminal state. The response may represent either success or failure; determine success by checking whether the error field exists. Examples:

Success response 200

{
  "status": "COMPLETED",
  "request_id": "mmsu_01hxyz...",
  "response_url": "https://.../response",
  "metrics": {
    "inference_time": 12.34
  }
}

Failure response 200

{
  "status": "COMPLETED",
  "request_id": "mmsu_01hxyz...",
  "response_url": "https://.../response",
  "error": "Task execution failed",
  "error_type": "INFERENCE_FAILURE"
}

Note

metrics.inference_time is platform-side observable duration. It may include queueing, scheduling, and result processing time, so do not treat it as pure model inference time.

Common error_type values:

error_type	Meaning
INFERENCE_FAILURE	Model execution failed
INFERENCE_TIMEOUT	Model execution timed out
REQUEST_INVALID	Request parameters are invalid or do not match model requirements
CONTENT_BLOCKED	Input or output content was blocked by safety policy
CANCELED	The task has been canceled
INTERNAL_SERVER_ERROR	Internal server error