Skip to main content

Reading a result

A successful API response and a useful answer are related, but they are not the same check. Read the transport status first, then evaluate the content against the purpose of your request.

In the web cabinet​

The answer is surrounded by the information needed to review it:

  • how OmevAI understood the request;
  • sources when live pages were used;
  • copy or download actions;
  • a positive or negative quality rating;
  • the matching line in Journal, with time, duration, status and cost.

If the journal shows a successful call but the content is not useful, improve the request or send precise feedback. If the journal shows a failure, keep the time and request identifier and use Errors and retries.

In an application​

Text calls use the OpenAI Chat Completions response shape:

{
  "id": "request-id",
  "model": "omev-pro",
  "choices": [{"index": 0, "message": {"role": "assistant", "content": "..."}}],
  "usage": {"prompt_tokens": 4123, "completion_tokens": 318, "total_tokens": 4441}
}
  • Read text from choices[0].message.content.
  • Keep the request identifier for logs and support.
  • When requesting JSON with response_format, parse it and validate it against your own schema.
  • Do not assume that an HTTP 200 proves factual or business quality. Review content according to the risk of the use case.

OmevAI can apply service-side safeguards and formal output checks where supported, but the ordinary text endpoint does not promise a separate human or schema review of every answer.

Cost and history​

Self-service users review calls in Dashboard, Usage, Journal and Billing in the web cabinet. Corporate users can additionally use the usage cabinet for organisation-wide reliability, invoices and exports.