Skip to main content
The API limits how many requests you can send in a window. When a request crosses a limit, the response says so, and the fix is to slow down and retry.

​
When you hit a limit

  • HTTP endpoints return 429 Too Many Requests.
  • WebSocket sessions send an error frame with the code rate_limit.

​
Handle it

Back off and retry. Space retries out with exponential backoff and jitter rather than retrying right away. When a response carries a Retry-After header, wait that many seconds before the next attempt. The agents control plane also sends Retry-After while a referenced tool or dependency is still preparing.

​
Send fewer requests

  • Move large, non-realtime audio to the Batch API instead of many single requests.
  • Keep one WebSocket session open for a stream instead of reconnecting per message.

​
Raise your limits

Contact support@slng.ai with your account and expected volume. See Error codes for the full error shape and WebSockets vs HTTP for choosing a protocol.