fix(detail): stabilize volcano multimodal plan/review image handling

app/api/canvas/detail-plan/route.ts

@@ -48,6 +48,14 @@ function supportsVisionPlanning(model: {
   );
 }
 
+function isVolcanoResponsesPlanner(config: {
+  provider?: { type?: string | null; baseUrl?: string | null } | null;
+}) {
+  if (!config?.provider || config.provider.type !== "custom") return false;
+  const baseUrl = String(config.provider.baseUrl || "").toLowerCase();
+  return baseUrl.includes("ark.cn-beijing.volces.com") || /\/api\/v3(\/responses)?\/?$/.test(baseUrl);
+}
+
 type DetailPlanRequest = {
   prompt?: string;
   count?: number;
@@ -1645,18 +1653,26 @@ export async function POST(request: Request) {
       .filter(Boolean)
       .join("\n");
 
-    const imageInputs =
+    let imageInputs: Array<{ type: "image"; image: string }> = [];
-      supportsVisionPlanning(plannerConfig.model) && referenceImages.length > 0
+    if (supportsVisionPlanning(plannerConfig.model) && referenceImages.length > 0) {
-        ? (
+      if (isVolcanoResponsesPlanner(plannerConfig)) {
-            await inlineVisionImages(referenceImages, {
+        imageInputs = referenceImages
-              maxImages: 6,
+          .slice(0, 6)
-              maxSide: 900,
+          .map((url) => ({ type: "image" as const, image: url }));
-              timeoutMs: 12000,
+      } else {
-              retries: 2,
+        const inlined = await inlineVisionImages(referenceImages, {
-              loggerPrefix: "[Canvas Detail Plan]",
+          maxImages: 6,
-            })
+          maxSide: 900,
-          ).inlineDataUrls.map((dataUrl) => ({ type: "image" as const, image: dataUrl }))
+          timeoutMs: 12000,
-        : [];
+          retries: 2,
+          loggerPrefix: "[Canvas Detail Plan]",
+        });
+        imageInputs = inlined.inlineDataUrls.map((dataUrl) => ({
+          type: "image" as const,
+          image: dataUrl,
+        }));
+      }
+    }
 
     let text = "";
     try {
@@ -1714,16 +1730,20 @@ export async function POST(request: Request) {
       } as any);
       text = String(result?.text || "");
     } catch (error) {
-      if (isAbortError(error)) {
+      const message = String((error as Error)?.message || "");
+      if (isAbortError(error) || /invalid scheme/i.test(message)) {
         console.error("[Canvas Detail Plan] Primary planning aborted, fallback applied:", error);
+        const fallbackReason = /invalid scheme/i.test(message)
+          ? "规划模型图片输入协议不兼容，已回退到稳定分镜模板。"
+          : "规划模型调用超时，已自动回退到稳定分镜模板。";
         return NextResponse.json({
           success: true,
           data: {
             ...fallback,
-            summary: `${fallback.summary}\n规划模型超时，已回退到稳定分镜模板。`,
+            summary: `${fallback.summary}\n${fallbackReason}`,
             qualityChecks: {
               passed: false,
-              issues: ["规划模型调用超时，已自动回退到稳定分镜模板。"],
+              issues: [fallbackReason],
             },
             model: {
               id: plannerConfig.model.id,
@@ -1734,7 +1754,7 @@ export async function POST(request: Request) {
               passed: false,
               designDirectorScore: 0,
               marketingDirectorScore: 0,
-              issues: ["规划阶段超时，已启用稳定模板回退。"],
+              issues: [fallbackReason],
               rewriteApplied: false,
             },
           },

app/api/canvas/detail-review/route.ts

@@ -13,7 +13,7 @@ import { getLanguageModelById } from "@/lib/ai/dynamic-provider";
 import { inlineVisionImages } from "@/lib/ai/vision-inline";
 
 export const runtime = "nodejs";
-export const maxDuration = 60;
+export const maxDuration = 180;
 
 type DetailReviewRequest = {
   outputLanguage?: string;
@@ -66,6 +66,33 @@ function supportsVisionReview(model: {
   );
 }
 
+function isVolcanoResponsesReviewer(config: {
+  provider?: { type?: string | null; baseUrl?: string | null } | null;
+}) {
+  if (!config?.provider || config.provider.type !== "custom") return false;
+  const baseUrl = String(config.provider.baseUrl || "").toLowerCase();
+  return baseUrl.includes("ark.cn-beijing.volces.com") || /\/api\/v3(\/responses)?\/?$/.test(baseUrl);
+}
+
+function isAbortError(error: unknown): boolean {
+  const target = error as {
+    name?: unknown;
+    message?: unknown;
+    code?: unknown;
+    ABORT_ERR?: unknown;
+  };
+  const name = String(target?.name || "");
+  const message = String(target?.message || "");
+  const code = Number(target?.code);
+  const abortCode = Number(target?.ABORT_ERR);
+  return (
+    name === "AbortError" ||
+    code === 20 ||
+    abortCode === 20 ||
+    /aborted|abort/i.test(message)
+  );
+}
+
 function normalizeLanguage(value: unknown): "zh" | "en" {
   const raw = String(value || "").toLowerCase();
   if (raw.includes("en") || raw.includes("english")) return "en";
@@ -246,35 +273,68 @@ export async function POST(request: Request) {
       .filter(Boolean)
       .join("\n");
 
-    const { inlineDataUrls } = await inlineVisionImages([imageUrl, ...referenceImages], {
+    const reviewImageCandidates = [imageUrl, ...referenceImages].slice(0, 6);
-      maxImages: 6,
+    let effectiveReviewImages = reviewImageCandidates;
-      maxSide: 900,
+    if (!isVolcanoResponsesReviewer(reviewerConfig)) {
-      timeoutMs: 12000,
+      const { inlineDataUrls } = await inlineVisionImages(reviewImageCandidates, {
-      retries: 2,
+        maxImages: 6,
-      loggerPrefix: "[Canvas Detail Review]",
+        maxSide: 900,
-    });
+        timeoutMs: 12000,
-    const effectiveReviewImages = inlineDataUrls.length > 0 ? inlineDataUrls : [imageUrl, ...referenceImages];
+        retries: 2,
+        loggerPrefix: "[Canvas Detail Review]",
+      });
+      effectiveReviewImages = inlineDataUrls.length > 0 ? inlineDataUrls : reviewImageCandidates;
+    }
 
     const content: Array<{ type: "text"; text: string } | { type: "image"; image: string }> = [
       { type: "text", text: promptParts },
       ...effectiveReviewImages.map((image) => ({ type: "image" as const, image })),
     ];
 
-    const { text } = await generateText({
+    let text = "";
-      model,
+    try {
-      messages: [
+      const result = await generateText({
-        {
+        model,
-          role: "user",
+        messages: [
-          content,
+          {
-        },
+            role: "user",
-      ],
+            content,
-      system:
+          },
-        language === "en"
+        ],
-          ? "You are both an Art Director and a Marketing Director for high-conversion e-commerce detail pages. Evaluate with strict professional standards and output JSON only."
+        system:
-          : "你同时是高转化电商详情页的平面设计总监与营销总监。按严格专业标准评估，仅输出JSON。",
+          language === "en"
-      temperature: 0.1,
+            ? "You are both an Art Director and a Marketing Director for high-conversion e-commerce detail pages. Evaluate with strict professional standards and output JSON only."
-      maxTokens: 800,
+            : "你同时是高转化电商详情页的平面设计总监与营销总监。按严格专业标准评估，仅输出JSON。",
-    } as any);
+        temperature: 0.1,
+        maxTokens: 800,
+      } as any);
+      text = String(result?.text || "");
+    } catch (error) {
+      const message = String((error as Error)?.message || "");
+      if (isAbortError(error) || /invalid scheme/i.test(message)) {
+        const fallbackIssues = isAbortError(error)
+          ? ["质检模型超时，已跳过二次审稿并保留本屏结果。"]
+          : ["质检模型图片输入协议不兼容，已跳过二次审稿并保留本屏结果。"];
+        return NextResponse.json({
+          success: true,
+          data: {
+            pass: true,
+            score: 78,
+            designDirectorScore: 78,
+            marketingDirectorScore: 78,
+            issues: fallbackIssues,
+            layoutFixes: [],
+            revisedPrompt: fallbackPrompt,
+            model: {
+              id: reviewerConfig.model.id,
+              displayName: reviewerConfig.model.displayName,
+              providerType: reviewerConfig.provider.type,
+            },
+          },
+        });
+      }
+      throw error;
+    }
 
     const parsed = extractJsonBlock<{
       pass?: unknown;

lib/ai/providers/volcano-multimodal.ts

@@ -130,8 +130,12 @@ async function callResponsesApi(params: {
   imageUrls: string[];
   timeout: number;
 }) {
+  const remoteImageUrls = params.imageUrls
+    .map((url) => String(url || "").trim())
+    .filter((url) => /^https?:\/\//i.test(url));
+
   const content = [
-    ...params.imageUrls.map((url) => ({ type: "input_image", image_url: url })),
+    ...remoteImageUrls.map((url) => ({ type: "input_image", image_url: url })),
     { type: "input_text", text: params.text || "请根据图片内容进行分析。" },
   ];
 
@@ -190,7 +194,7 @@ async function callResponsesApi(params: {
 
 export function createVolcanoMultimodal(config: VolcanoMultimodalConfig) {
   const endpoint = normalizeResponsesEndpoint(config.baseURL);
-  const timeout = config.timeout ?? 60000;
+  const timeout = config.timeout ?? 120000;
   const toV2Usage = (usage: { promptTokens?: number; completionTokens?: number }) => {
     const inputTokens = usage.promptTokens ?? 0;
     const outputTokens = usage.completionTokens ?? 0;