feat(web): 模型能力 - Agnes 图像协议补充比例档位与单次张数默认值
This commit is contained in:
@@ -236,6 +236,26 @@ export function defaultImageCapabilityConfig(protocol?: ModelProtocol, model = "
|
||||
image.outputFormat = { supported: false };
|
||||
image.maxOutputs = 4;
|
||||
}
|
||||
if (protocol === "agnes-image") {
|
||||
// Agnes 图像:size 必填,取 1K/2K/3K/4K 档位或 WxH 精确尺寸,画面比例走独立的 ratio 字段;
|
||||
// 参考图放 extra_body.image,支持多图合成,但没有蒙版端点。
|
||||
image.references.maxImages = 9;
|
||||
image.references.maskSupported = false;
|
||||
image.size = {
|
||||
parameter: "aspect_ratio",
|
||||
values: ["1:1", "3:4", "4:3", "16:9", "9:16", "2:3", "3:2", "21:9"],
|
||||
default: "1:1",
|
||||
allowCustom: false,
|
||||
};
|
||||
// 官方档位是 1K/2K/3K/4K,统一层只提供 1k/2k/4k 三档;3K 保留在协议映射里但不在界面露出。
|
||||
image.quality = { supported: true, values: ["1k", "2k", "4k"], default: "2k" };
|
||||
image.transparentBackground = { supported: false, default: false };
|
||||
// 顶层 response_format 是官方明确的错误写法,输出格式只能在 extra_body 内声明。
|
||||
image.responseFormat = { supported: false };
|
||||
image.outputFormat = { supported: false };
|
||||
// 该端点不接受 n,单次请求固定返回一张图片。
|
||||
image.maxOutputs = 1;
|
||||
}
|
||||
if (protocol !== "grok-image" && model.trim().toLowerCase().startsWith("grok-imagine-image")) {
|
||||
image.references.maxImages = 0;
|
||||
image.references.maskSupported = false;
|
||||
|
||||
Reference in New Issue
Block a user