Kiểm tra Prompt
Kiểm tra văn bản, hình ảnh hoặc cả hai trước khi tạo, và nhận kết quả cho phép, gắn cờ hoặc từ chối cùng điểm số cho từng danh mục moderation.
Ủy quyền
Bearer authentication header of the form Bearer <token>, where <token> is your auth token.
Nội dung
A screen of text, an image, or both. Send at least one of text and image.
The image to screen, as base64, with or without a data:image/...;base64, prefix. The
formats are JPEG, PNG, WebP, GIF and BMP. The limit is 6991530 base64 characters, and the
decoded image must be at most 5 MiB.
6991530Your identifier for this screen, up to 128 characters, with no control characters. The
response returns it in request_id.
The text to screen, up to 8000 characters.
8000Phản hồi
The verdict.
The verdict of one screen.
The probability, from 0 to 1, that the screen falls in each category.
True when real-person likeness and sexual content together crossed their combined threshold, the pattern of a sexual deepfake.
The verdict. allow means the content passed. deny means block the content. flag means
apply your own judgement. It is not a soft deny.
allow, flag, deny The time the screen took, in milliseconds.
x >= 0True when the text was also screened in a normalized form, with obfuscation such as invisible or look-alike characters removed.
Human-readable reasons for the decision. The wording can change, so do not parse it.
The number of yes/no questions the model answered for this screen.
x >= 0How each score in categories was measured.
The request_id you sent, or null.
The categories whose score crossed the threshold of the category. It can be empty on a
flag from the general check. notes then gives the reason.
A moderation category.
violent_crimes, sex_related_crimes, child_sexual_exploitation, suicide_and_self_harm, indiscriminate_weapons, intellectual_property, defamation, non_violent_crimes, hate, privacy, specialized_advice, sexual_content, non_consensual_intimate_imagery, minor_coded_language, real_person_likeness, living_artist_style, prompt_injection