Skip to main content

​
How It Works

Every call to POST /api/v1/run/{model_id} returns immediately with a task_id: 
{
  "task_id": "f47ac10b-58cc-4372-a567-0e02b2c3d479",
  "status": "waiting"
}
The actual generation happens in the background. You retrieve the result by either polling or using a callback URL.

​
Task Lifecycle

waiting → generating → success
                     → fail
StatusMeaning
waitingTask queued, not yet started
generatingModel is actively generating
successGeneration complete — result_urls is populated
failGeneration failed — check error_message

​
Option 1 — Polling

Call GET /api/v1/run/tasks/{task_id} repeatedly until status is success or fail. 
curl https://routerbase.kuaibaike.com/api/v1/run/tasks/f47ac10b-58cc-4372-a567-0e02b2c3d479 \
  -H "Authorization: Bearer sk-t0-xxxxxxxxxxxx"
Recommended polling interval: 3 seconds for video/image, 1 second for chat models.

​
Option 2 — Callback URL

Pass a callback_url in your request body. RouterBase will POST the completed task result to that URL. 
{
  "prompt": "A sunset over the ocean",
  "callback_url": "https://yourapp.com/webhooks/routerbase",
  "video_params": { "aspect_ratio": "16:9" }
}
The callback payload matches the same structure as the poll response: 
{
  "task_id": "f47ac10b-58cc-4372-a567-0e02b2c3d479",
  "status": "success",
  "result_urls": ["https://routerbase.kuaibaike.com/results/video.mp4"],
  "error_message": null,
  "created_at": "2025-01-15T10:30:00Z"
}
Callbacks eliminate the need for polling and reduce latency for long-running tasks like video generation (typically 1–3 minutes).

​
Typical Generation Times

ModalityTypical Time
Chat / LLM2–30 seconds
Image generation10–60 seconds
Video generation1–5 minutes
Times vary by model and provider load.