48.3 Error patterns and what they usually mean

Overview and links for this section of the guide.

On this page

Errors

  • 429 Too Many Requests: You are sending messages too fast. Implement exponential backoff (wait 1s, then 2s, then 4s).
  • 400 Bad Request (Safety): You triggered a safety filter (Sex/Hate/Harassment). Check your input data.
  • 500 Internal Server Error: Google/OpenAI is down. Retry once, then give up.
  • Output truncated: You hit the `max_output_tokens` limit. Increase the limit or ask the model to be brief.

Where to go next