latest2026.01.16 DevInference Service

Vision Task

VisionTask does vision analysis (e.g., image captioning or recognition) within a session. The request must include the session identifier and image data.

HTTP Mapping: POST “/v1/conversation/inference/vision”

Request

This endpoint expects an object.
namestringRequired
The resource name of the session.
imagestringRequiredformat: "bytes"
The image data to be analyzed.

Response

OK
captionstringRead-only
The resulting caption or analysis from the vision model.

Errors

400
Bad Request Error