Integration guide
Handle API errors
Keep the ad request independent from your AI workflow. A rejected request or an unavailable ad should leave the model response and result accessible.
Read the response before deciding what to do
Check the HTTP status and error code. Authentication or configuration failures need a corrected request; repeating the same request does not fix them. Use the error-code reference for the response envelope and code definitions.
A successful response with decision.fill: false is a no-fill outcome, not a transport failure. Hide the empty slot and continue the user flow. Do not turn no-fill into an automatic retry loop.
429 rate-limit response
Use the Retry-After header before retrying a rate_limited response.
429-rate-limited.http
text
Respect the placement window
Any retry must still fit your application's timing. If the result is ready or the placement window has ended, stop requesting a waiting-state ad. Keep network failures and timeouts out of the model's response path.
Keep diagnostics useful and private
Use the request identifier and error code to find the event in your logs. Do not include secret keys, raw prompts or full user conversations in browser logs or support messages. See troubleshooting for integration checks.
Need help with your integration?
Share the affected endpoint, request ID, and the behavior you expected. Leave out keys and user content.