aisplice

When something goes wrong

FAQ & support

#Common questions

Which model should I use?

Start with the model each capability page recommends; it is the best value among the models online today. Models & Pricing lists every model with its live price, and switching is one field in the request.

Does the gateway support tool calling / function calling?

Yes, for every OpenAI-shaped model plus Claude and Gemini. Send the standard tools array; the gateway forwards it and returns tool_calls in the response. Streaming tool-call deltas are supported.

Can I use it in Cursor, Cherry Studio or other apps?

Yes. Open “Use it in an app” in Getting started, pick your app and copy its settings — the address, your key and a model name. No code needed.

How is quota deducted?

Pre-reserved at request start (worst-case estimate for token models, sticker price for task models), reconciled to the actual usage the vendor reports when the response finishes. If the request fails, the reservation is refunded automatically.

What if I hit 401 or 402?

401 means the bearer token is wrong or the key was deleted / disabled. 402 means the wallet balance cannot cover the pre-reserved estimate — top up and retry. Neither is a bug in the model.

Can I use image / video / music generation the same way?

Image generation uses the async /v1/images/tasks flow. Video uses POST /v1/videos, and music uses POST /v1/music/generations. All async task routes support submit → poll → content, and can optionally deliver a webhook when the task reaches a terminal state.

Are my prompts logged?

Chat calls: only when a call fails do we keep a trimmed copy of the request and reply (7 days, to look into the failure), unless you turn on “Record request content” on a key. Image, video and music requests keep a trimmed copy for 7 days. Metadata (model, tokens, cost, time, request id) is kept 30 days. Only our operations team can see any of it.

How long is my data kept?

Generated images, videos and music files: 3 days, so download what you want to keep. Trimmed request copies for troubleshooting: 7 days. Request records on the Usage page, image, video and music tasks, and notifications: 30 days; the Billing page lists the same 30 days. Sign-in records: 180 days. Billing records — every top-up, charge and refund — and daily totals are kept permanently: export any period of up to 31 days from the Billing page as a CSV.

Can I rate-limit or scope a key per project?

Yes. Each key can carry its own RPM, RPD, quota cap, model whitelist, IP allowlist, and group. Rotate anything from Console → Keys without touching the code.

#Support

Something not working, or a model you want us to add? Two ways to reach us.

Still stuck?

Send us the request ID or the error message and we will look into it.