11

2026-04-16 21:47:20 +08:00
parent 2456c10c63
commit 09d4c0c31b
79 changed files with 11873 additions and 2341 deletions
--- a/server-node/src/services/sceneImageService.ts
+++ b/server-node/src/services/sceneImageService.ts
@@ -19,6 +19,8 @@ export const sceneImageSchema = z.object({
  landmarkId: z.string().trim().optional().default(''),
  referenceImageSrc: z.string().trim().optional().default(''),
 });
+const TEXT_TO_IMAGE_SCENE_MODEL = 'wan2.2-t2i-flash';
+const REFERENCE_IMAGE_SCENE_MODEL = 'qwen-image-2.0';

 function parseImageDataUrl(source: string) {
  const matched = /^data:(image\/[^;]+);base64,(.+)$/u.exec(source);
@@ -122,40 +124,72 @@ function extractImageUrls(payload: Record<string, unknown>) {
  return [...new Set(urls)];
 }

-function ensurePayload(
-  payload: z.infer<typeof sceneImageSchema>,
-  defaultModel: string,
-) {
-  if (!payload.landmarkName && !payload.landmarkId) {
-    throw badRequest('landmarkName 或 landmarkId 至少要提供一个');
+async function createSceneImageTask(params: {
+  baseUrl: string;
+  apiKey: string;
+  payload: z.infer<typeof sceneImageSchema>;
+}) {
+  const { baseUrl, apiKey, payload } = params;
+  const response = await fetch(`${baseUrl}/services/aigc/image-generation/generation`, {
+    method: 'POST',
+    headers: {
+      Authorization: `Bearer ${apiKey}`,
+      'Content-Type': 'application/json',
+      'X-DashScope-Async': 'enable',
+    },
+    body: JSON.stringify({
+      model: payload.model,
+      input: {
+        messages: [
+          {
+            role: 'user',
+            content: [{ text: payload.prompt }],
+          },
+        ],
+      },
+      parameters: {
+        n: 1,
+        size: payload.size,
+        prompt_extend: true,
+        watermark: false,
+        ...(payload.negativePrompt
+          ? { negative_prompt: payload.negativePrompt }
+          : {}),
+      },
+    }),
+  });
+  const responseText = await response.text();
+
+  if (!response.ok) {
+    return {
+      ok: false as const,
+      errorMessage: extractApiErrorMessage(
+        responseText,
+        '创建场景图片生成任务失败',
+      ),
+    };
  }

  return {
-    ...payload,
-    model: payload.model || defaultModel,
+    ok: true as const,
+    payload: JSON.parse(responseText) as Record<string, unknown>,
  };
 }

-export async function generateSceneImage(
-  context: AppContext,
-  input: z.infer<typeof sceneImageSchema>,
-) {
-  const payload = ensurePayload(input, context.config.dashScope.imageModel);
-  const baseUrl = context.config.dashScope.baseUrl.replace(/\/+$/u, '');
-  const referenceImage = payload.referenceImageSrc
-    ? await resolveReferenceImageAsDataUrl(
-        context.config.projectRoot,
-        payload.referenceImageSrc,
-      )
-    : '';
-  const createResponse = await fetch(
-    `${baseUrl}/services/aigc/image-generation/generation`,
+async function createSceneImageFromReference(params: {
+  baseUrl: string;
+  apiKey: string;
+  payload: z.infer<typeof sceneImageSchema>;
+  referenceImage: string;
+}) {
+  const { baseUrl, apiKey, payload, referenceImage } = params;
+  const response = await fetch(
+    `${baseUrl}/services/aigc/multimodal-generation/generation`,
    {
      method: 'POST',
      headers: {
-        Authorization: `Bearer ${context.config.dashScope.apiKey}`,
+        Authorization: `Bearer ${apiKey}`,
        'Content-Type': 'application/json',
-        'X-DashScope-Async': 'enable',
      },
      body: JSON.stringify({
        model: payload.model,
@@ -163,10 +197,7 @@ export async function generateSceneImage(
          messages: [
            {
              role: 'user',
-              content: [
-                { text: payload.prompt },
-                ...(referenceImage ? [{ image: referenceImage }] : []),
-              ],
+              content: [{ image: referenceImage }, { text: payload.prompt }],
            },
          ],
        },
@@ -182,56 +213,65 @@ export async function generateSceneImage(
      }),
    },
  );
-  const createText = await createResponse.text();
-  if (!createResponse.ok) {
-    throw badRequest(
-      extractApiErrorMessage(createText, '创建场景图片生成任务失败'),
-    );
-  }
-
-  const createPayload = JSON.parse(createText) as Record<string, unknown>;
-  const taskId = extractTaskId(createPayload);
-  if (!taskId) {
-    throw badRequest('场景图片生成任务未返回 task_id');
-  }
-
-  const deadline = Date.now() + context.config.dashScope.requestTimeoutMs;
-  let imageUrl = '';
-  let actualPrompt = '';
-
-  while (Date.now() < deadline) {
-    const pollResponse = await fetch(`${baseUrl}/tasks/${taskId}`, {
-      headers: {
-        Authorization: `Bearer ${context.config.dashScope.apiKey}`,
-      },
-    });
-    const pollText = await pollResponse.text();
-    if (!pollResponse.ok) {
-      throw badRequest(
-        extractApiErrorMessage(pollText, '查询场景图片任务失败'),
-      );
-    }
-
-    const pollPayload = JSON.parse(pollText) as Record<string, unknown>;
-    const status = findFirstStringByKey(pollPayload, 'task_status').trim();
-    if (status === 'SUCCEEDED') {
-      imageUrl = extractImageUrls(pollPayload)[0] ?? '';
-      actualPrompt = findFirstStringByKey(pollPayload, 'actual_prompt').trim();
-      break;
-    }
-    if (status === 'FAILED' || status === 'UNKNOWN') {
-      throw badRequest(
-        extractApiErrorMessage(pollText, '场景图片生成任务失败'),
-      );
-    }
-
-    await new Promise((resolve) => setTimeout(resolve, 2000));
+  const responseText = await response.text();
+
+  if (!response.ok) {
+    return {
+      ok: false as const,
+      errorMessage: extractApiErrorMessage(
+        responseText,
+        '创建参考图场景编辑任务失败',
+      ),
+    };
  }

+  const responsePayload = JSON.parse(responseText) as Record<string, unknown>;
+  const imageUrl = extractImageUrls(responsePayload)[0] ?? '';
  if (!imageUrl) {
-    throw badRequest('场景图片生成超时或未返回图片地址');
+    return {
+      ok: false as const,
+      errorMessage: '参考图场景编辑未返回图片地址',
+    };
  }

+  return {
+    ok: true as const,
+    imageUrl,
+    actualPrompt: findFirstStringByKey(responsePayload, 'actual_prompt').trim(),
+    taskId: `scene-edit-${Date.now()}`,
+  };
+}
+
+function ensurePayload(
+  payload: z.infer<typeof sceneImageSchema>,
+  _defaultModel: string,
+) {
+  if (!payload.landmarkName && !payload.landmarkId) {
+    throw badRequest('landmarkName 或 landmarkId 至少要提供一个');
+  }
+
+  const referenceImageSrc =
+    typeof payload.referenceImageSrc === 'string'
+      ? payload.referenceImageSrc.trim()
+      : '';
+
+  return {
+    ...payload,
+    referenceImageSrc,
+    model: referenceImageSrc
+      ? REFERENCE_IMAGE_SCENE_MODEL
+      : TEXT_TO_IMAGE_SCENE_MODEL,
+  };
+}
+
+async function saveSceneImageAsset(params: {
+  context: AppContext;
+  payload: z.infer<typeof sceneImageSchema>;
+  imageUrl: string;
+  taskId: string;
+  actualPrompt: string;
+}) {
+  const { context, payload, imageUrl, taskId, actualPrompt } = params;
  const imageResponse = await fetch(imageUrl);
  if (!imageResponse.ok) {
    throw badRequest('下载生成图片失败');
@@ -295,3 +335,99 @@ export async function generateSceneImage(
    actualPrompt,
  };
 }
+
+export async function generateSceneImage(
+  context: AppContext,
+  input: z.infer<typeof sceneImageSchema>,
+) {
+  const payload = ensurePayload(input, context.config.dashScope.imageModel);
+  const baseUrl = context.config.dashScope.baseUrl.replace(/\/+$/u, '');
+  const referenceImage = payload.referenceImageSrc.trim()
+    ? await resolveReferenceImageAsDataUrl(
+        context.config.projectRoot,
+        payload.referenceImageSrc,
+      )
+    : '';
+
+  if (referenceImage) {
+    const referenceResult = await createSceneImageFromReference({
+      baseUrl,
+      apiKey: context.config.dashScope.apiKey,
+      payload,
+      referenceImage,
+    });
+
+    if (!referenceResult.ok) {
+      throw badRequest(referenceResult.errorMessage);
+    }
+
+    return saveSceneImageAsset({
+      context,
+      payload,
+      imageUrl: referenceResult.imageUrl,
+      taskId: referenceResult.taskId,
+      actualPrompt: referenceResult.actualPrompt,
+    });
+  }
+
+  const createTaskResult = await createSceneImageTask({
+    baseUrl,
+    apiKey: context.config.dashScope.apiKey,
+    payload,
+  });
+
+  if (!createTaskResult.ok) {
+    throw badRequest(createTaskResult.errorMessage);
+  }
+
+  const createPayload = createTaskResult.payload;
+  const taskId = extractTaskId(createPayload);
+  if (!taskId) {
+    throw badRequest('场景图片生成任务未返回 task_id');
+  }
+
+  const deadline = Date.now() + context.config.dashScope.requestTimeoutMs;
+  let imageUrl = '';
+  let actualPrompt = '';
+
+  while (Date.now() < deadline) {
+    const pollResponse = await fetch(`${baseUrl}/tasks/${taskId}`, {
+      headers: {
+        Authorization: `Bearer ${context.config.dashScope.apiKey}`,
+      },
+    });
+    const pollText = await pollResponse.text();
+    if (!pollResponse.ok) {
+      throw badRequest(
+        extractApiErrorMessage(pollText, '查询场景图片任务失败'),
+      );
+    }
+
+    const pollPayload = JSON.parse(pollText) as Record<string, unknown>;
+    const status = findFirstStringByKey(pollPayload, 'task_status').trim();
+    if (status === 'SUCCEEDED') {
+      imageUrl = extractImageUrls(pollPayload)[0] ?? '';
+      actualPrompt = findFirstStringByKey(pollPayload, 'actual_prompt').trim();
+      break;
+    }
+    if (status === 'FAILED' || status === 'UNKNOWN') {
+      throw badRequest(
+        extractApiErrorMessage(pollText, '场景图片生成任务失败'),
+      );
+    }
+
+    await new Promise((resolve) => setTimeout(resolve, 2000));
+  }
+
+  if (!imageUrl) {
+    throw badRequest('场景图片生成超时或未返回图片地址');
+  }
+
+  return saveSceneImageAsset({
+    context,
+    payload,
+    imageUrl,
+    taskId,
+    actualPrompt,
+  });
+}