Submit managed text and embedding requests

Before you start

Have a customer API key, selected account, sufficient wallet funds, and an enabled text or embedding model.

Steps

1. Read the model catalog through the playground or ispeed-ctl models list. Note the model ID, supported operation, limits, and prices.
2. Prepare input and an appropriate maximum output limit. Embedding requests use the catalog’s embedding operation and no text-generation output budget.
3. Use the quote operation for an estimate where available. An estimate does not guarantee live capacity.
4. Submit once with a unique idempotency key, then poll or wait using the returned request ID.
5. Read the completed result and measured token usage. When retrying the same submission, reuse its idempotency key only with the exact same payload.

Example

ispeed-ctl models list
ispeed-ctl requests submit --body-file request.json --idempotency-key text-example-0001
ispeed-ctl requests wait REQUEST_ID --timeout 10m

Check the result

The request completes and actual usage settles against its reservation; unused reserved funds are released.

Troubleshooting

A disabled model, insufficient funds, or exceeded limit should be resolved before resubmitting. A poll timeout does not authorize creating a second request.