UlazAI developer docs
Errors and limits
Error handling and platform limits
Use structured handling for API errors and validate inputs client-side to reduce retries.
Common HTTP responses
| Code | Meaning | Typical cause |
|---|---|---|
400 |
Invalid request | Missing prompt, unsupported size, invalid model params |
401 |
Unauthorized | Missing or invalid API key/session token |
402 |
Payment required | Insufficient purchased credits or API access restriction |
403 |
Forbidden | API register/login endpoints are intentionally disabled |
404 |
Not found | Generation ID not owned by current user or endpoint mismatch |
500 |
Server error | Upstream provider failure or network timeout |
Request limits to enforce client-side
prompt: maximum 5000 characters on unified endpoint.image_urls: maximum 14 URLs on unified endpoint.size: model-specific, validate against model matrix.quality: model-specific values for Seedream 5 Lite, Nano Banana 2, Ideogram, and Qwen.- API usage requires purchased credits; welcome credits are website-only.
Retry strategy
Recommended production behavior:
- Do not retry
400,401,402, or403without changing input or auth. - Retry
500with exponential backoff and idempotency keying in your app layer. - Poll status endpoint before launching duplicate generation jobs.