图片顺序改变问题处理

src/app/api/generate/providers/__tests__/popiserver.test.ts

@@ -35,6 +35,10 @@ function makeInput(): GenerationInput {
     },
     prompt: "@角色1 和 @角色2 在一起吃火锅",
     images: ["https://example.com/1.png", "https://example.com/2.png"],
+    referenceSubjectList: [
+      { id: 1, type: "image", url: "https://example.com/1.png", name: "图1" },
+      { id: 2, type: "image", url: "https://example.com/2.png", name: "图2" },
+    ],
     subType: POPI_VIDEO_SUBTYPE_OMNI_REFERENCE,
     parameters: {
       ratio: "16:9",
@@ -97,15 +101,11 @@ describe("popiserver generation provider", () => {
     expect(body).toMatchObject({
       type: 2,
       subType: POPI_VIDEO_SUBTYPE_OMNI_REFERENCE,
-      aiPlatform: "GATEWAY",
       origin: "canvas",
-      model: "",
       aiModelId: 15,
       aiModelCode: "viduq2-pro",
       aiModelCodeAlias: "viduq2-pro",
-      batchSize: 1,
       ratio: "16:9",
-      aspect_ratio: "16:9",
       resolution: "720",
       videoRatio: 10,
     });
@@ -197,12 +197,48 @@ describe("popiserver generation provider", () => {
       },
       referenceSubjectList: [
         { id: 1, type: "image", url: "https://example.com/1.png", name: "图1" },
-        { id: 2, type: "image", url: "https://example.com/2.png", name: "图2" },
       ],
     });
     expect(body.generate_audio).toBeUndefined();
   });
 
+  it("uses the frontend referenceSubjectList order without rebuilding it from images", async () => {
+    vi.mocked(global.fetch).mockResolvedValue({
+      ok: true,
+      json: () => Promise.resolve({
+        status: "0000",
+        message: "ok",
+        data: { id: 1198 },
+      }),
+    } as Response);
+
+    const input = makeInput();
+    input.images = [
+      "https://popitest-public-1313913486.cos.ap-guangzhou.myqcloud.com/media/image/2026/0702/9535.jpeg",
+      "https://popitest-public-1313913486.cos.ap-guangzhou.myqcloud.com/media/image/2026/0701/9225.png",
+    ];
+    input.referenceSubjectList = [
+      {
+        id: 1,
+        type: "image",
+        url: "https://popitest-public-1313913486.cos.ap-guangzhou.myqcloud.com/media/image/2026/0701/9225.png",
+        name: "图1",
+      },
+      {
+        id: 2,
+        type: "image",
+        url: "https://popitest-public-1313913486.cos.ap-guangzhou.myqcloud.com/media/image/2026/0702/9535.jpeg",
+        name: "图2",
+      },
+    ];
+
+    await submitPopiTask(makeRequest(), "login-token", input);
+
+    const [, init] = vi.mocked(global.fetch).mock.calls[0];
+    const body = JSON.parse(String(init?.body));
+    expect(body.referenceSubjectList).toEqual(input.referenceSubjectList);
+  });
+
   it("collects all task ids from batched task create responses", async () => {
     vi.mocked(global.fetch).mockResolvedValue({
       ok: true,
@@ -239,10 +275,7 @@ describe("popiserver generation provider", () => {
     const [, init] = vi.mocked(global.fetch).mock.calls[0];
     const body = JSON.parse(String(init?.body));
     expect(body).toMatchObject({
-      ratio: "9:16",
       aspect_ratio: "9:16",
-      width: 720,
-      height: 1280,
     });
     expect(body.aspectRatio).toBeUndefined();
   });
@@ -283,7 +316,6 @@ describe("popiserver generation provider", () => {
     const body = JSON.parse(String(init?.body));
     expect(body).toMatchObject({
       chatPrompt: "比例：16:9\n清晰度：1K\n提示词：create a cinematic cat",
-      model: "",
       ratio: "16:9",
       resolution: "1K",
     });
@@ -325,15 +357,13 @@ describe("popiserver generation provider", () => {
     const body = JSON.parse(String(init?.body));
     expect(body).toMatchObject({
       chatPrompt: "比例：9:16\n清晰度：2K\n提示词：create a clean product render",
-      model: "",
       ratio: "9:16",
-      aspect_ratio: "9:16",
       resolution: "2K",
     });
     expect(body.aspectRatio).toBeUndefined();
   });
 
-  it("builds referenceSubjectList for image tasks without @ in names", async () => {
+  it("uses frontend referenceSubjectList for image tasks", async () => {
     vi.mocked(global.fetch).mockResolvedValue({
       ok: true,
       json: () => Promise.resolve({
@@ -359,18 +389,18 @@ describe("popiserver generation provider", () => {
     };
     input.prompt = "@图1 作为人物参考生成海报";
     input.images = ["https://statictest.popi.art/media/ani/character/2026/0512/4540.jpg"];
-    input.subType = 103;
+    input.referenceSubjectList = [
-    input.parameters = {
-      ratio: "16:9",
-      resolution: "1K",
-      referenceSubjectList: [
       {
         id: 1,
         type: "image",
-          url: "stale-image-url",
+        url: "https://statictest.popi.art/media/ani/character/2026/0512/4540.jpg",
-          name: "@图1",
+        name: "图1",
       },
-      ],
+    ];
+    input.subType = 103;
+    input.parameters = {
+      ratio: "16:9",
+      resolution: "1K",
     };
 
     const result = await submitPopiTask(makeRequest(), "login-token", input);
@@ -410,15 +440,14 @@ describe("popiserver generation provider", () => {
     expect(JSON.parse(String(init?.body))).toEqual({
       type: 3,
       chatPrompt: "我是一个小白兔，白白的小白兔",
-      model: "",
       origin: "canvas",
-      aiPlatform: "GATEWAY",
       aiModelName: "Speech 2.8 HD",
       aiModelId: 30,
       aiModelCode: "speech-2.8-hd",
       aiModelCodeAlias: "speech-2.8-hd",
       subType: 301,
-      batchSize: 1,
+      referenceSubjectList: undefined,
+      extraTaskParams: {},
     });
   });
 
@@ -585,7 +614,7 @@ describe("popiserver generation provider", () => {
     ]);
   });
 
-  it("builds referenceSubjectList from final image and video URLs", async () => {
+  it("uses frontend referenceSubjectList with image, video, and audio references", async () => {
     vi.mocked(global.fetch).mockResolvedValue({
       ok: true,
       json: () => Promise.resolve({
@@ -612,39 +641,39 @@ describe("popiserver generation provider", () => {
     };
     input.prompt = "@图1的人物参考@视频2的人物的动作跳舞";
     input.images = ["https://statictest.popi.art/media/ani/character/2026/0512/4540.jpg"];
-    input.dynamicInputs = {
+    input.referenceSubjectList = [
-      videos: ["https://statictest.popi.art/media/aiGen/2026/0603/5470.mp4"],
-      voices: ["https://statictest.popi.art/media/audio/2026/0603/5470.mp3"],
-    };
-    input.parameters = {
-      aiModelName: "seedance 2.0-fast",
-      aiPlatform: "seedance 2.0",
-      model: "seedance 2.0",
-      ratio: "16:9",
-      resolution: "480P",
-      duration: 5,
-      batchSize: 1,
-      referenceSubjectList: [
       {
         id: 1,
         type: "image",
-          url: "stale-image-url",
+        url: "https://statictest.popi.art/media/ani/character/2026/0512/4540.jpg",
-          name: "@图1",
+        name: "图1",
       },
       {
         id: 2,
         type: "video",
-          url: "stale-video-url",
+        url: "https://statictest.popi.art/media/aiGen/2026/0603/5470.mp4",
-          name: "@视频2",
+        name: "视频2",
         duration: 5.033,
       },
       {
         id: 3,
         type: "audio",
-          url: "stale-audio-url",
+        url: "https://statictest.popi.art/media/audio/2026/0603/5470.mp3",
-          name: "@音频3",
+        name: "音频3",
       },
-      ],
+    ];
+    input.dynamicInputs = {
+      videos: ["https://statictest.popi.art/media/aiGen/2026/0603/5470.mp4"],
+      voices: ["https://statictest.popi.art/media/audio/2026/0603/5470.mp3"],
+    };
+    input.parameters = {
+      aiModelName: "seedance 2.0-fast",
+      aiPlatform: "seedance 2.0",
+      model: "seedance 2.0",
+      ratio: "16:9",
+      resolution: "480P",
+      duration: 5,
+      batchSize: 1,
     };
     input.subType = POPI_VIDEO_SUBTYPE_OMNI_REFERENCE;
 
@@ -666,33 +695,11 @@ describe("popiserver generation provider", () => {
       resolution: "480P",
       duration: 5,
       batchSize: 1,
-      width: 1280,
-      height: 720,
       chatPrompt: "@图1的人物参考@视频2的人物的动作跳舞",
       images: ["https://statictest.popi.art/media/ani/character/2026/0512/4540.jpg"],
       videos: ["https://statictest.popi.art/media/aiGen/2026/0603/5470.mp4"],
       voices: ["https://statictest.popi.art/media/audio/2026/0603/5470.mp3"],
-      referenceSubjectList: [
+      referenceSubjectList: input.referenceSubjectList,
-        {
-          id: 1,
-          type: "image",
-          url: "https://statictest.popi.art/media/ani/character/2026/0512/4540.jpg",
-          name: "图1",
-        },
-        {
-          id: 2,
-          type: "video",
-          url: "https://statictest.popi.art/media/aiGen/2026/0603/5470.mp4",
-          name: "视频2",
-          duration: 5.033,
-        },
-        {
-          id: 3,
-          type: "audio",
-          url: "https://statictest.popi.art/media/audio/2026/0603/5470.mp3",
-          name: "音频3",
-        },
-      ],
     });
   });
 
@@ -714,43 +721,13 @@ describe("popiserver generation provider", () => {
     input.dynamicInputs = {
       videos: ["https://statictest.popi.art/media/aiGen/2026/0603/5470.mp4"],
     };
-    input.parameters = {
-      ...input.parameters,
-      referenceSubjectList: [
-        {
-          id: 1,
-          type: "image",
-          url: "stale-image-url",
-          name: "@图1",
-        },
-        {
-          id: 2,
-          type: "video",
-          url: "stale-video-url",
-          name: "@视频2",
-        },
-      ],
-    };
 
     await submitPopiTask(makeRequest(), "login-token", input);
 
     const [, init] = vi.mocked(global.fetch).mock.calls[0];
     const body = JSON.parse(String(init?.body));
     expect(body.videos).toBeUndefined();
-    expect(body.referenceSubjectList).toEqual([
+    expect(body.referenceSubjectList).toEqual(input.referenceSubjectList);
-      {
-        id: 1,
-        type: "image",
-        url: "https://example.com/1.png",
-        name: "图1",
-      },
-      {
-        id: 2,
-        type: "image",
-        url: "https://example.com/2.png",
-        name: "图2",
-      },
-    ]);
   });
 
   it("extracts completed image task output from task list", async () => {

src/utils/popiserverTaskPayload.ts

@@ -8,14 +8,6 @@ export type PopiModelDetailForTaskPayload = {
   aiModelCodeAlias?: string;
 };
 
-type ReferenceSubject = {
-  id: number;
-  type: "image" | "video" | "audio";
-  url: string;
-  name: string;
-  duration?: number;
-};
-
 const POPISERVER_LEGACY_AUDIO_KEYS = [
   "audio",
   "audios",
@@ -98,39 +90,6 @@ function buildPopiChatPrompt(
   ].join("\n");
 }
 
-function normalizeReferenceSubjectName(name: string): string {
-  return name.replace(/^@+/, "");
-}
-
-function referenceSubjectMetadata(referenceSubjectList: unknown): ReferenceSubject[] {
-  if (!Array.isArray(referenceSubjectList)) return [];
-
-  return referenceSubjectList.flatMap((item) => {
-    if (!item || typeof item !== "object" || Array.isArray(item)) return [];
-    const subject = item as Record<string, unknown>;
-    const type =
-      subject.type === "video"
-        ? "video"
-        : subject.type === "image"
-          ? "image"
-          : subject.type === "audio"
-            ? "audio"
-            : null;
-    if (!type) return [];
-    const id = asNumber(subject.id);
-    const name = asString(subject.name);
-    const url = asString(subject.url);
-    const duration = asNumber(subject.duration);
-    return [{
-      id: id ?? 0,
-      type,
-      url: url ?? "",
-      name: normalizeReferenceSubjectName(name ?? ""),
-      ...(duration !== null ? { duration } : {}),
-    }];
-  });
-}
-
 function removeLegacyAudioParameters(parameters: Record<string, unknown>): Record<string, unknown> {
   return Object.fromEntries(
     Object.entries(parameters).filter(
@@ -139,57 +98,6 @@ function removeLegacyAudioParameters(parameters: Record<string, unknown>): Recor
   );
 }
 
-function buildReferenceSubjectList(
-  images: string[],
-  videos: string[],
-  voices: string[],
-  referenceSubjectList: unknown
-): ReferenceSubject[] {
-  const metadata = referenceSubjectMetadata(referenceSubjectList);
-  const metadataByType = {
-    image: metadata.filter((item) => item.type === "image"),
-    video: metadata.filter((item) => item.type === "video"),
-    audio: metadata.filter((item) => item.type === "audio"),
-  };
-  const subjects: ReferenceSubject[] = [];
-
-  images.forEach((url, index) => {
-    const id = subjects.length + 1;
-    const meta = metadataByType.image[index];
-    subjects.push({
-      id,
-      type: "image",
-      url,
-      name: normalizeReferenceSubjectName(meta?.name || `图${id}`),
-    });
-  });
-
-  videos.forEach((url, index) => {
-    const id = subjects.length + 1;
-    const meta = metadataByType.video[index];
-    subjects.push({
-      id,
-      type: "video",
-      url,
-      name: normalizeReferenceSubjectName(meta?.name || `视频${id}`),
-      ...(meta?.duration ? { duration: meta.duration } : {}),
-    });
-  });
-
-  voices.forEach((url, index) => {
-    const id = subjects.length + 1;
-    const meta = metadataByType.audio[index];
-    subjects.push({
-      id,
-      type: "audio",
-      url,
-      name: normalizeReferenceSubjectName(meta?.name || `音频${id}`),
-    });
-  });
-
-  return subjects;
-}
-
 export function buildPopiTaskBody(
   input: GenerationInput,
   modelDetail: PopiModelDetailForTaskPayload | null
@@ -197,7 +105,7 @@ export function buildPopiTaskBody(
   const rawParameters = removeLegacyAudioParameters(input.parameters || {});
   const rawExtraTaskParams = removeLegacyAudioParameters(input.extraTaskParams || {});
   const aspectRatio = resolveAspectRatio(rawParameters);
-  const { referenceSubjectList: _referenceSubjectList, ...parameters } = rawParameters;
+  const parameters = rawParameters;
   const selectedModel = {
     provider: "popiserver" as const,
     modelId: input.model.id,
@@ -226,9 +134,6 @@ export function buildPopiTaskBody(
   const voices = modelSupportsReferenceMedia(selectedModel, "audio", rawParameters)
     ? input.voices || []
     : [];
-  const referenceSubjectList = type === 1 || type === 2
-    ? buildReferenceSubjectList([...new Set(images)], [...new Set(videos)], [...new Set(voices)], input.referenceSubjectList)
-    : [];
   const voiceId =
     asString(input.audioVoiceId) ??
     asString(parameters.voiceId) ??
@@ -249,7 +154,7 @@ export function buildPopiTaskBody(
     ...(asNumber(input.batchSize) !== null || asNumber(parameters.batchSize) !== null
       ? { batchSize: asNumber(input.batchSize) ?? asNumber(parameters.batchSize) }
       : {}),
-    referenceSubjectList,
+    referenceSubjectList: input.referenceSubjectList,
     extraTaskParams: rawExtraTaskParams,
   };