U9 AI · DOCUMENTATION
Troubleshooting & support
Diagnose authentication, model routing, streaming and missing output.
Download Markdown ↓Diagnose in order#
| Symptom | Check |
|---|---|
| 401 / 403 | Correct site key, active key, expiry, key restrictions and correct authentication header |
| Model unavailable | Exact model ID, account access, enabled routing and endpoint protocol |
| Insufficient credit | Available versus reserved credit and the selected generation specification |
| 400 / 422 | Model-specific required fields, supported duration, aspect ratio and reference-image format |
| 429 | Rate limit or upstream capacity; honor Retry-After and avoid request bursts |
| 5xx / timeout | Save request ID and time; check existing task before resubmitting |
| Agent can chat but tools fail | Responses / Messages support, streaming events, tool-call round trips and client version |
| Completed without download | Task details, output URL and resource expiry; report the original task |
Useful support ticket#
Open a private ticket with the model ID, generation mode, request/task ID, exact time with time zone, client version, endpoint path, HTTP status and sanitized error text. State expected and actual behavior. Attach a redacted screenshot if helpful. Never attach a complete Authorization header, API key, session cookie, private key or unredacted config file.
Avoid duplicate charges#
A network timeout can happen after the upstream accepted a job. Query existing task details first. Do not repeatedly click Generate or automatically retry creation unless the model's contract explicitly guarantees idempotency. Use request history, wallet history and cost analysis together to establish the final result and charge.
← Documentation