fix(agent): enrich generic image-analysis text with visual summary

app/(chat)/api/chat/route.ts

@@ -798,6 +798,45 @@ function buildAutoModeFallbackMessage(params: {
   ].join("\n");
 }
 
+function extractMessageImageUrls(parts: any[]): string[] {
+  return (parts || [])
+    .map((part: any) => {
+      if (!part) return null;
+      if (part.type === "file") return part.url || part.data || null;
+      if (part.type === "image") return part.url || part.image || null;
+      if (part.type === "image_url") return part.image_url?.url || null;
+      return null;
+    })
+    .filter((url: any): url is string => typeof url === "string" && /^https?:\/\//i.test(url));
+}
+
+function shouldEnrichImageAnalysisText(message?: string): boolean {
+  if (!message) return false;
+  if (!/###\s*图片分析/.test(message)) return false;
+  return /(基于当前上传图片处理|基于你选中的原图处理|基于上一轮图片继续处理)/.test(message);
+}
+
+function injectImageAnalysisSummary(message: string, summary: string): string {
+  const normalizedSummary = summary.replace(/\s+/g, " ").trim();
+  if (!normalizedSummary) return message;
+  const lines = message.split("\n");
+  const analysisHeaderIndex = lines.findIndex((line) => /^###\s*图片分析/.test(line.trim()));
+  if (analysisHeaderIndex < 0) return message;
+
+  for (let i = analysisHeaderIndex + 1; i < lines.length; i += 1) {
+    const current = lines[i]?.trim() || "";
+    if (/^###\s*/.test(current)) break;
+    if (/^- /.test(current)) {
+      lines[i] = `- ${normalizedSummary}`;
+      return lines.join("\n");
+    }
+  }
+
+  const insertIndex = analysisHeaderIndex + 1;
+  lines.splice(insertIndex, 0, `- ${normalizedSummary}`);
+  return lines.join("\n");
+}
+
 /**
  * 自定义消息转换函数
  * 注意：gemini-image 和 huanke-gemini 都需要 URL 格式（不是 Base64）
@@ -2099,6 +2138,44 @@ export async function POST(request: Request) {
       });
     }
 
+    if (
+      shouldRunAgentConversation &&
+      !smartAutoNeedsClarification &&
+      smartAutoFriendlyMessage &&
+      shouldEnrichImageAnalysisText(smartAutoFriendlyMessage)
+    ) {
+      try {
+        const currentImageUrls = extractMessageImageUrls(message.parts as any[]);
+        const summaryImageUrls =
+          currentImageUrls.length > 0 ? currentImageUrls : smartAutoHistoryImages;
+
+        if (summaryImageUrls.length > 0) {
+          const { summarizeImageForExecutionReport } = await import("@/lib/ai/smart-analyzer");
+          const summary = await withOperationTimeout(
+            summarizeImageForExecutionReport({
+              text: message.parts
+                .filter((part): part is { type: "text"; text: string } => part.type === "text")
+                .map((part) => part.text)
+                .join(" "),
+              imageUrls: summaryImageUrls,
+              context: typeof autoContext === "string" ? autoContext : undefined,
+            }),
+            "enrichImageAnalysisSummary",
+            8000,
+          );
+          if (summary?.trim()) {
+            smartAutoFriendlyMessage = injectImageAnalysisSummary(
+              smartAutoFriendlyMessage,
+              summary.trim(),
+            );
+            console.log("[Smart Auto Mode] ✅ Enriched image analysis summary");
+          }
+        }
+      } catch (enrichError) {
+        console.warn("[Smart Auto Mode] Skip image analysis enrichment:", enrichError);
+      }
+    }
+
     const session = await auth();
 
     if (!session?.user) {

lib/ai/smart-analyzer.ts

@@ -636,6 +636,47 @@ export async function summarizeImageForExecutionReport(params: {
     if (summary) return summary;
   }
 
+  try {
+    const selectedVisionModel =
+      (agentModel && isLikelyVisionCapableModel(agentModel) ? agentModel : undefined) ||
+      allModels.find((m) => isLikelyVisionCapableModel(m));
+    if (!selectedVisionModel) return null;
+
+    const { getLanguageModelById } = await import("./dynamic-provider");
+    const model = await getLanguageModelById(selectedVisionModel.model.id);
+    if (!model) return null;
+
+    const fallbackVisionImages = await buildAnalyzerVisionImagesByMode({
+      imageUrls,
+      preferRemoteUrls: shouldPreferRemoteVisionUrls(selectedVisionModel),
+    });
+    if (fallbackVisionImages.length === 0) return null;
+
+    const { text: summaryText } = await generateText({
+      model,
+      messages: [
+        {
+          role: "user",
+          content: [
+            {
+              type: "text",
+              text: `${promptText}\n只输出一行，格式：画面要点：主体+场景+风格+光线（不超过45字）。`,
+            },
+            ...fallbackVisionImages,
+          ],
+        },
+      ],
+      temperature: 0.1,
+      maxTokens: 120,
+      abortSignal: AbortSignal.timeout(8000),
+    } as any);
+
+    const normalized = summaryText?.replace(/\s+/g, " ").trim();
+    return normalized || null;
+  } catch (error) {
+    console.warn("[Smart Analyzer] summarizeImageForExecutionReport fallback failed:", error);
+  }
+
   return null;
 }