feat(platform): 创作能力整合 - 新增 Gemini Images 图片请求协议

2026-08-18 21:42:22 +08:00
parent a23b4dda93
commit 778551127c
17 changed files with 489 additions and 74 deletions
--- a/web/src/lib/gemini-image.ts
+++ b/web/src/lib/gemini-image.ts
@@ -0,0 +1,40 @@
+export type GeminiImageGenerationConfig = {
+    responseModalities: ["TEXT", "IMAGE"];
+    imageConfig?: { aspectRatio?: string; imageSize?: "1K" | "2K" | "4K" };
+};
+
+export type GeminiInlineImage = {
+    mimeType: string;
+    data: string;
+};
+
+const GEMINI_IMAGE_SIZES: Record<string, GeminiInlineImageGenerationSize> = {
+    low: "1K",
+    medium: "2K",
+    high: "4K",
+    "1k": "1K",
+    "2k": "2K",
+    "4k": "4K",
+};
+
+type GeminiInlineImageGenerationSize = "1K" | "2K" | "4K";
+
+export function parseGeminiImageDataUrl(dataUrl: string): GeminiInlineImage {
+    const match = dataUrl.trim().match(/^data:([^;,\s]+)(?:;[^,]*)?;base64,([A-Za-z0-9+/=_-]+)$/i);
+    if (!match) throw new Error("参考图片读取失败：未得到有效的 image data URL");
+    const mimeType = match[1].toLowerCase();
+    const data = match[2].replace(/-/g, "+").replace(/_/g, "/");
+    if (!mimeType.startsWith("image/")) throw new Error(`参考图片 MIME 类型无效：${mimeType}`);
+    if (!data) throw new Error("参考图片读取失败：图片数据为空");
+    return { mimeType, data };
+}
+
+export function buildGeminiImageGenerationConfig(size?: string, quality?: string): GeminiImageGenerationConfig {
+    const imageConfig: NonNullable<GeminiImageGenerationConfig["imageConfig"]> = {};
+    const normalizedSize = size?.trim();
+    const normalizedQuality = quality?.trim().toLowerCase();
+    if (normalizedSize && normalizedSize !== "auto" && /^\d+:\d+$/.test(normalizedSize)) imageConfig.aspectRatio = normalizedSize;
+    const imageSize = GEMINI_IMAGE_SIZES[normalizedQuality || ""];
+    if (imageSize) imageConfig.imageSize = imageSize;
+    return { responseModalities: ["TEXT", "IMAGE"], ...(Object.keys(imageConfig).length ? { imageConfig } : {}) };
+}
--- a/web/src/lib/model-capabilities.ts
+++ b/web/src/lib/model-capabilities.ts
@@ -77,10 +77,45 @@ export type VideoCapabilityConfig = {
 // Keep explicit pixel presets for each resolution tier so the settings panel can
 // switch between 1K, 2K and 4K without silently converting the requested ratio.
 const defaultImageSizes = [
-    "auto", "1:1", "3:2", "2:3", "4:3", "3:4", "16:9", "21:9", "9:16",
-    "1024x1024", "1360x1024", "1024x1360", "1536x1024", "1024x1536", "1024x1280", "1280x1024", "2048x878", "1824x1024", "1024x1824",
-    "2048x2048", "2304x1728", "1728x2304", "2496x1664", "1664x2496", "1792x2240", "2240x1792", "3136x1344", "2752x1536", "1536x2752",
-    "2880x2880", "3264x2448", "2448x3264", "3504x2336", "2336x3504", "2560x3200", "3200x2560", "3808x1632", "3840x2160", "2160x3840",
+    "auto",
+    "1:1",
+    "3:2",
+    "2:3",
+    "4:3",
+    "3:4",
+    "16:9",
+    "21:9",
+    "9:16",
+    "1024x1024",
+    "1360x1024",
+    "1024x1360",
+    "1536x1024",
+    "1024x1536",
+    "1024x1280",
+    "1280x1024",
+    "2048x878",
+    "1824x1024",
+    "1024x1824",
+    "2048x2048",
+    "2304x1728",
+    "1728x2304",
+    "2496x1664",
+    "1664x2496",
+    "1792x2240",
+    "2240x1792",
+    "3136x1344",
+    "2752x1536",
+    "1536x2752",
+    "2880x2880",
+    "3264x2448",
+    "2448x3264",
+    "3504x2336",
+    "2336x3504",
+    "2560x3200",
+    "3200x2560",
+    "3808x1632",
+    "3840x2160",
+    "2160x3840",
 ];

 export function defaultImageCapabilityConfig(protocol?: ModelProtocol, model = ""): ImageCapabilityConfig {
@@ -123,6 +158,20 @@ export function defaultImageCapabilityConfig(protocol?: ModelProtocol, model = "
        image.responseFormat.supported = false;
        image.outputFormat.supported = false;
    }
+    if (protocol === "gemini-image") {
+        image.references.maskSupported = false;
+        // Gemini Images uses imageConfig.aspectRatio, not the OpenAI-style pixel size field.
+        image.size = {
+            parameter: "aspect_ratio",
+            values: ["auto", "1:1", "2:3", "3:2", "3:4", "4:3", "9:16", "16:9", "21:9"],
+            default: "1:1",
+            allowCustom: false,
+        };
+        image.transparentBackground = { supported: false, default: false };
+        image.responseFormat = { supported: false };
+        image.outputFormat = { supported: false };
+        image.maxOutputs = 4;
+    }
    if (protocol !== "grok-image" && model.trim().toLowerCase().startsWith("grok-imagine-image")) {
        image.references.maxImages = 0;
        image.references.maskSupported = false;
@@ -207,20 +256,14 @@ export function modelCapabilityConfigFor(config: { channels: Array<{ id: string;
    const cost = channel?.modelCosts?.find((item) => item.model === modelName);
    const fallback = defaultModelCapabilityConfig(cost?.protocol, modelName);
    if (!cost?.capabilityConfig) return fallback;
-    const text = cost.capabilityConfig.text
-        ? { ...fallback.text!, ...cost.capabilityConfig.text, references: { ...fallback.text!.references, ...cost.capabilityConfig.text.references } }
-        : fallback.text;
-    const video = cost.capabilityConfig.video
-        ? { ...fallback.video!, ...cost.capabilityConfig.video, references: { ...fallback.video!.references, ...cost.capabilityConfig.video.references } }
-        : fallback.video;
+    const text = cost.capabilityConfig.text ? { ...fallback.text!, ...cost.capabilityConfig.text, references: { ...fallback.text!.references, ...cost.capabilityConfig.text.references } } : fallback.text;
+    const video = cost.capabilityConfig.video ? { ...fallback.video!, ...cost.capabilityConfig.video, references: { ...fallback.video!.references, ...cost.capabilityConfig.video.references } } : fallback.video;
    return { ...fallback, ...cost.capabilityConfig, text, image: cost.capabilityConfig.image || fallback.image, video };
 }

 export function normalizeImageValue(profile: ImageCapabilityConfig, value: { size?: string; quality?: string; count?: string; transparentBackground?: string }) {
    const size = normalizeImageSizeSetting(profile, value.size);
-    const quality = profile.quality.supported
-        ? (value.quality && profile.quality.values.includes(value.quality) ? value.quality : profile.quality.default || "auto")
-        : profile.quality.default || "auto";
+    const quality = profile.quality.supported ? (value.quality && profile.quality.values.includes(value.quality) ? value.quality : profile.quality.default || "auto") : profile.quality.default || "auto";
    const count = String(Math.max(1, Math.min(profile.maxOutputs, Math.floor(Math.abs(Number(value.count)) || 1))));
    const transparentBackground = profile.transparentBackground.supported && value.transparentBackground === "true" ? "true" : "false";
    return { size, quality, count, transparentBackground };
@@ -242,16 +285,16 @@ export function imageSizeRequest(profile: ImageCapabilityConfig, value?: string)
 }

 export function normalizeVideoValue(profile: VideoCapabilityConfig, value: { seconds?: string; ratio?: string; resolution?: string }) {
-    const duration = profile.duration.selection === "enum"
-        ? (profile.duration.values || []).includes(Number(value.seconds)) ? Number(value.seconds) : profile.duration.default
-        : normalizeRangeDuration(profile, Number(value.seconds));
+    const duration = profile.duration.selection === "enum" ? ((profile.duration.values || []).includes(Number(value.seconds)) ? Number(value.seconds) : profile.duration.default) : normalizeRangeDuration(profile, Number(value.seconds));
    const ratio = profile.ratios.includes(value.ratio || "") ? value.ratio! : profile.defaultRatio;
    const resolution = profile.resolutions.includes(value.resolution || "") ? value.resolution! : profile.defaultResolution;
    return { seconds: String(duration), ratio, resolution };
 }

 export function videoResolutionRequest(profile: VideoCapabilityConfig, value: string | undefined) {
-    const requested = String(value || "").trim().toLowerCase();
+    const requested = String(value || "")
+        .trim()
+        .toLowerCase();
    if (!requested || requested === "auto" || requested === "default" || requested === "medium" || requested === "high") return undefined;
    const candidates = [requested];
    if (/^\d+$/.test(requested)) candidates.push(`${requested}p`);
--- a/web/src/lib/model-protocols.ts
+++ b/web/src/lib/model-protocols.ts
@@ -5,6 +5,7 @@ export type ModelProtocol =
    | "grok-image"
    | "volcengine-ark-image"
    | "volcengine-jimeng-image"
+    | "gemini-image"
    | "openai-audio"
    | "async-audio"
    | "newapi"
@@ -35,6 +36,7 @@ export const MODEL_PROTOCOLS: ModelProtocolDefinition[] = [
    { value: "grok-image", label: "Grok Images", capability: "image", create: "POST /v1/images/generations / edits", contentType: "application/json", media: "文生图与单张 URL 参考图，不支持蒙版" },
    { value: "volcengine-ark-image", label: "火山方舟图片", capability: "image", create: "POST /api/v3/images/generations", contentType: "application/json", media: "文生图与 image 参考图，不支持蒙版" },
    { value: "volcengine-jimeng-image", label: "即梦官方图片", capability: "image", create: "POST CVSync2AsyncSubmitTask", poll: "POST CVSync2AsyncGetResult", contentType: "application/json + AK/SK 签名", media: "0-14 张参考图，模型标识填写 req_key" },
+    { value: "gemini-image", label: "Gemini Images", capability: "image", create: "POST /v1beta/models/{model}:generateContent", contentType: "application/json", media: "文生图、图片编辑与多张内嵌参考图" },
    { value: "openai-audio", label: "OpenAI Audio", capability: "audio", create: "POST /v1/audio/speech", contentType: "application/json", media: "文本转语音" },
    { value: "async-audio", label: "异步音频任务", capability: "audio", create: "POST /v1/audio/tasks", poll: "GET /v1/audio/tasks/{task_id}", contentType: "application/json", media: "语音、音效与音乐生成" },
    { value: "newapi", label: "OpenAI / NewAPI Videos", capability: "video", create: "POST /v1/videos", poll: "GET /v1/videos/{task_id}", contentType: "multipart/form-data", media: "input_reference[] 参考图" },
@@ -96,6 +98,7 @@ export function protocolForModelCatalog(endpointTypes: string[] = []): ModelProt
    const normalized = new Set(endpointTypes.map((value) => value.trim().toLowerCase()));
    if (normalized.has("openai-chat") || normalized.has("chat-completion") || normalized.has("chat")) return "chat-completion";
    if (normalized.has("openai-response") || normalized.has("responses")) return "openai-response";
+    if (normalized.has("gemini-image") || normalized.has("gemini-images")) return "gemini-image";
    if (normalized.has("openai-image") || normalized.has("image")) return "openai-image";
    if (normalized.has("openai-video") || normalized.has("video")) return "newapi-channel-2";
    if (normalized.has("openai-audio") || normalized.has("audio")) return "openai-audio";