diff --git a/app/(chat)/api/chat/route.ts b/app/(chat)/api/chat/route.ts
index 13e41177c..15429a18e 100644
--- a/app/(chat)/api/chat/route.ts
+++ b/app/(chat)/api/chat/route.ts
@@ -696,11 +696,7 @@ export async function POST(request: Request) {
         const needsRicherAutoReply =
           isImageTask &&
           !smartAutoNeedsClarification &&
-          (
-            !smartAutoFriendlyMessage ||
-            !smartAutoFriendlyMessage.includes("\n") ||
-            smartAutoFriendlyMessage.replace(/\s+/g, "").length < 20
-          );
+          !smartAutoFriendlyMessage;
 
         if (needsRicherAutoReply) {
           smartAutoFriendlyMessage = buildAutoModeFallbackMessage({
diff --git a/lib/ai/auto-agent-graph.ts b/lib/ai/auto-agent-graph.ts
index 354ec0ed8..da6054068 100644
--- a/lib/ai/auto-agent-graph.ts
+++ b/lib/ai/auto-agent-graph.ts
@@ -1,4 +1,5 @@
 import { Annotation, StateGraph, START, END } from "@langchain/langgraph";
+import { generateText } from "ai";
 import type { AnalysisResult } from "@/lib/ai/smart-analyzer";
 import type { RouteResult } from "@/lib/ai/model-router";
 
@@ -47,38 +48,87 @@ function compactText(text: string, max = 48) {
 function buildStrategicFriendlyMessage(
   analysis: AnalysisResult,
   state: AutoAgentStateType,
-): string | null {
+): Promise<string | null> {
   const fallbackQuestion = analysis.clarifyingQuestions?.[0];
 
   if (analysis.needsClarification) {
-    return analysis.conversationalReply || fallbackQuestion || "我先确认一下，你希望我优先调整哪一部分？";
+    return Promise.resolve(
+      analysis.conversationalReply || fallbackQuestion || "我先确认一下，你希望我优先调整哪一部分？"
+    );
   }
 
-  if (analysis.intent === "image_gen") {
-    const demand = compactText(state.text);
-    return [
-      "收到，我来帮你走自动模式生成方案。",
-      `- 需求理解：${demand}`,
-      "- 执行策略：先锁定主体与场景，再控制风格/光影/构图，最后输出可继续迭代的版本。",
-      "- 继续优化：你可以直接说“保留A，改B”，我会按最新要求快速迭代。",
-    ].join("\n");
-  }
+  return generateAgentConversation(analysis, state);
+}
 
-  if (analysis.intent === "image_edit") {
-    const sourceHint = state.hasImage
-      ? state.imageFromHistory
-        ? "我会基于你上一轮图片继续编辑。"
-        : "我会基于你刚上传的图片编辑。"
-      : "我会先按当前上下文进行编辑。";
-    return [
-      "收到，我来帮你精准编辑这张图。",
-      `- 执行基准：${sourceHint}`,
-      "- 处理方式：只改你提到的目标区域，其他主体、构图与风格尽量保持不变。",
-      "- 交付方式：先出一版结果，你再告诉我细调方向（如颜色、材质、氛围、细节）。",
-    ].join("\n");
-  }
+async function generateAgentConversation(
+  analysis: AnalysisResult,
+  state: AutoAgentStateType
+): Promise<string | null> {
+  try {
+    const { getEnabledAiModelsWithProvider } = await import("@/lib/db/queries");
+    const { getLanguageModelById } = await import("./dynamic-provider");
+    const allModels = await getEnabledAiModelsWithProvider();
 
-  return analysis.conversationalReply || null;
+    const candidateModel = state.hasImage
+      ? allModels.find((m) => m.model.supportsVision && m.model.isEnabled)
+      : allModels.find((m) => m.model.isEnabled && !m.model.isSubModel);
+
+    if (!candidateModel) {
+      return analysis.conversationalReply || null;
+    }
+
+    const model = await getLanguageModelById(candidateModel.model.id);
+    if (!model) {
+      return analysis.conversationalReply || null;
+    }
+
+    const history = state.historyMessages
+      .slice(-6)
+      .map((msg) => `${msg.role === "user" ? "用户" : "助手"}: ${compactText(msg.content, 120)}`)
+      .join("\n");
+
+    const imageInfo = state.hasImage
+      ? `用户本轮可用图片数量：${state.imageUrls.length || 1}。图片来源：${state.imageFromHistory ? "历史消息复用" : "本轮上传"}。`
+      : "用户本轮没有上传图片。";
+
+    const systemPrompt = `你是一个真实执行中的 AI 生图/修图 Agent。
+请根据用户需求和分析结果，给出“正在执行前”的简短汇报。
+
+要求：
+1) 不要说“我无法看到图片”；按系统提供的信息做合理分析。
+2) 必须看起来是本轮真实分析结果，避免空泛套话。
+3) 不暴露内部实现、模型路由、算法细节。
+4) 输出 4-7 行 Markdown，语气专业自然。
+5) 如果是图像任务，包含以下小节标题：需求理解、分析结论、执行计划、交付说明。
+6) 每一行尽量具体，不要重复。`;
+
+    const prompt = [
+      `【用户原始需求】${state.text || "（无文本）"}`,
+      `【意图分析】intent=${analysis.intent}; subIntent=${analysis.subIntent || "none"}; confidence=${analysis.confidence}`,
+      `【图片信息】${imageInfo}`,
+      state.context ? `【画布上下文】${compactText(state.context, 300)}` : "",
+      history ? `【最近对话】\n${history}` : "",
+      "请输出本轮执行前汇报：",
+    ]
+      .filter(Boolean)
+      .join("\n\n");
+
+    const { text } = await generateText({
+      model,
+      system: systemPrompt,
+      prompt,
+      temperature: 0.35,
+      maxTokens: 420,
+    } as any);
+
+    const cleaned = text.trim();
+    if (cleaned.length < 10) {
+      return analysis.conversationalReply || null;
+    }
+    return cleaned;
+  } catch (error) {
+    return analysis.conversationalReply || null;
+  }
 }
 
 const analyzeNode = async (state: AutoAgentStateType) => {
@@ -96,7 +146,7 @@ const analyzeNode = async (state: AutoAgentStateType) => {
     ? analysis.clarifyingQuestions?.[0]
     : undefined;
 
-  const strategicMessage = buildStrategicFriendlyMessage(analysis, state);
+  const strategicMessage = await buildStrategicFriendlyMessage(analysis, state);
 
   return {
     analysis,