FAQ
Platform & account
How is SDKMAX different from going directly to OpenAI/Claude/Gemini? SDKMAX is an aggregation layer: one account, one API key, one OpenAI-compatible protocol reaches many providers — avoiding separate signups, keys, and bills per vendor, while adding automatic channel failover. See Platform Overview.
Do I need a business entity to sign up? No — individual developers can register and use it directly. For team/enterprise collaboration features, check the team settings after logging in, or contact the platform team to get one enabled.
Forgot your password? Use the "forgot password" flow on the sign-in page; if self-service reset isn't available yet, contact the platform team.
API Keys
Is an API key the same as my account password? No. Your password logs into the web console; the API key (sk- prefix) authenticates programmatic calls — they're independent and carry different privileges. See Create an API Key.
Can one key be scoped to specific models/groups? Yes, at creation/edit time you can restrict a key to a group or model allowlist, useful for cost or access isolation.
My key leaked — what now? Disable or delete it immediately in the console, issue a replacement, and check that key's usage history for anything unexpected.
Making calls
Why does the official OpenAI SDK just work with SDKMAX? Because SDKMAX's protocol matches OpenAI's — change base_url to https://api.sdkmax.com/v1 and swap in an SDKMAX key, and the rest of your integration is unchanged. See OpenAI-Compatible API.
How do I know which models are available right now? Call GET /v1/models, or check the console's model page. The list syncs automatically with upstream channels — don't hardcode a static "all available models" list in your business logic.
Is streaming (stream: true) supported? Yes, identical to OpenAI's SSE streaming protocol — see Making Requests.
Does it support function/tool calling? Yes, as long as the underlying model supports it — request/response fields follow the OpenAI shape.
When is the Batch API launching? Not yet — see Batch API for a concurrency-based interim approach. This page will be updated once it's officially supported.
Errors & rate limits
I'm getting 429 — what do I do? That means you've hit a quota cap or rate limit. Retry with exponential backoff and check that key's quota settings. See Error Codes.
I'm getting model_not_found — what do I do? Call GET /v1/models to confirm the model is currently available — when an upstream channel deprecates a model, the gateway removes it too. Avoid hardcoding model names without a fallback/degradation path.
My image/video job is stuck at "processing" — what now? Generation jobs (especially video) can genuinely take a while — poll at the recommended interval. If it's stuck well beyond the model's normal duration, contact the platform team to check channel status.
Billing & quota
Is quota tracked per key or per account? Both — the account has an overall quota, and individual keys can have their own cap; whichever is stricter applies. Check the console for exact numbers.
Where do I see detailed usage? Log in to the console's Usage/Billing page, broken down by key and by model.
Still stuck?
- Check the relevant page first: Platform Overview · Making Requests · Error Codes
- Still unresolved? Reach the platform team through the console's feedback/support channel, and include the request ID, timestamp, model, and which key was involved (no need to share the full secret) to speed up debugging.