Skip to main content
POST

Authorizations

string
required
API key, Bearer Token formatGet API Key:Visit the API Key management page to get your API Key

Body

string
required
Model name. Fixed value jev.The result.result.model in the response is the actual version number used upstream, such as jev-1.14.0. This field is returned by the upstream service; do not provide a version number in the request.
object
required
Judgment content. Contains state and questions, plus assets when images are included.
A single request can contain multiple noul, choice, score, and box questions simultaneously. Do not pass stream; this endpoint only returns a single complete JSON response.

Response

The response body is the raw JSON from upstream, which this endpoint returns without modification. The evaluation results are in result.result.answers.
boolean
Whether the request was successful
array
Error list. Empty array on success
array
Additional messages. Empty array when there are no messages
object
Upstream execution result

Usage Examples

Multi-Question Example

choice and score can be included in the same request as noul.

图片与 Box 示例(Beta)

Images are placed in assets, and state points to them using a reference name; `photo` is used in questions to refer to this image. The following example asks both “whether there is a car in the image” and “where the car is”.
The answers in the response:
Box questions always return a box, and there is no result such as “not found”; a box will still be provided even when there is no target in the image, only with lower confidence. When unsure whether the target exists, add a noul question in the same request as shown in the example above; ignore the box if noul is very low.

Image Requirements

  • Scale the long edge down to within 1568 pixels before sending. Upstream will strip EXIF data and will not rotate according to EXIF orientation; please orient the pixels of phone photos properly beforehand.
  • Images will be converted into input tokens, counted towards usage.input_tokens, and also occupy the context window. An image of about 800 pixels is approximately 1500 tokens.
  • If any image in the request fails to meet the requirements, the entire request fails, and partial results will not be returned.

注意事项

  • This endpoint is not a chat endpoint and does not accept messages or stream.
  • Image input and Box questions are currently in Beta. Request formats, results, and limitations may change across versions, and they are not recommended for direct use in production decisions.
  • The context window is 32000 tokens.
  • When a business failure occurs, the HTTP status code may still be 200. Please check both success and errors. Authentication failure returns 401, a missing model or invalid JSON returns 400, and a non-existent model returns 404.