fix(ssa): Fix Agent mode chat leaking plan/result details + navigation reset

Backend (ChatHandlerService):
- Replace LLM-generated hints with fixed-text hints for plan/code/result steps
  to prevent SystemPrompt intent instruction conflict causing verbose chat output
- Add sendFixedHint() helper for deterministic SSE text delivery (no LLM call)
- Cancel old plan_pending/code_pending executions when regenerating plan
- Eliminates 3 unnecessary LLM calls per analysis cycle (faster response)

Frontend (SSAModule):
- Use location.key as useEffect dependency to ensure store reset on every
  navigation (fixes stale session when re-entering from other modules)
- TopNavigation uses replace:true for active module re-click to avoid
  browser history clutter

Tested: Agent mode plan/code/result hints now show brief fixed text in chat,
detailed content exclusively in right workspace panel

Made-with: Cursor

backend/src/modules/ssa/services/ChatHandlerService.ts

@@ -478,6 +478,31 @@ export class ChatHandlerService {
     return 'other';
   }
 
+  /**
+   * 发送固定文本引导语（不经过 LLM，避免指令冲突导致 LLM 输出冗长内容）
+   */
+  private async sendFixedHint(
+    writer: StreamWriter,
+    placeholderMessageId: string,
+    text: string,
+  ): Promise<void> {
+    const sseData = JSON.stringify({
+      id: `chatcmpl-ssa-${Date.now()}`,
+      object: 'chat.completion.chunk',
+      choices: [{ delta: { content: text }, finish_reason: null }],
+    });
+    writer.write(`data: ${sseData}\n\n`);
+
+    const doneData = JSON.stringify({
+      id: `chatcmpl-ssa-${Date.now()}`,
+      object: 'chat.completion.chunk',
+      choices: [{ delta: {}, finish_reason: 'stop' }],
+    });
+    writer.write(`data: ${doneData}\n\n`);
+
+    await conversationService.finalizeAssistantMessage(placeholderMessageId, text);
+  }
+
   // ── Agent Step 1: 生成分析计划 → 等用户确认 ──
 
   private async agentGeneratePlan(
@@ -491,6 +516,12 @@ export class ChatHandlerService {
       writer.write(`data: ${JSON.stringify({ type, ...data })}\n\n`);
     };
 
+    // 取消该 session 下所有未完成的 pending 执行（防止旧计划干扰）
+    await (prisma as any).ssaAgentExecution.updateMany({
+      where: { sessionId, status: { in: ['plan_pending', 'code_pending'] } },
+      data: { status: 'error', errorMessage: '用户发起了新的分析请求' },
+    });
+
     const execution = await (prisma as any).ssaAgentExecution.create({
       data: { sessionId, query: userContent, status: 'planning' },
     });
@@ -516,17 +547,9 @@ export class ChatHandlerService {
       planText: plan.rawText,
     });
 
-    // 简短视线牵引语（方案 B：左侧只输出简短提示，详细内容在右侧工作区）
-    const hint = [
-      `[系统指令——严格遵守] 分析计划已生成（${plan.steps.length} 步），标题「${plan.title}」。`,
-      '你只需回复 1-2 句话，引导用户去右侧工作区查看。',
-      '示例回复：「我已为您拟定了详细的分析计划，请在右侧工作区核对并点击确认。」',
-      '【铁律】禁止在对话中重复计划内容、禁止编造任何数值。回复不超过 2 句话。',
-    ].join('\n');
-
-    const msgs = await conversationService.buildContext(sessionId, conversationId, 'analyze', hint);
-    const sr = await conversationService.streamToSSE(msgs, writer, { temperature: 0.3, maxTokens: 150 });
-    await conversationService.finalizeAssistantMessage(placeholderMessageId, sr.content, sr.thinking, sr.tokens);
+    // 固定文本引导语（不经过 LLM，彻底避免 SystemPrompt 的 intent 指令冲突）
+    const hintText = `我已为您拟定了分析计划（${plan.steps.length} 步），👉 请在右侧工作区核对并点击确认。`;
+    await this.sendFixedHint(writer, placeholderMessageId, hintText);
 
     return { messageId: placeholderMessageId, intent: 'analyze', success: true };
   }
@@ -571,17 +594,9 @@ export class ChatHandlerService {
       explanation: generated.explanation,
     });
 
-    // 简短视线牵引语
-    const hint = [
-      `[系统指令——严格遵守] R 代码已生成（${generated.code.split('\n').length} 行），使用 ${generated.requiredPackages.join(', ') || '基础包'}。`,
-      '你只需回复 1-2 句话，引导用户去右侧工作区核对代码并点击执行。',
-      '示例回复：「R 代码已生成，请在右侧工作区核对代码并点击 "确认并执行"。」',
-      '【铁律】禁止在对话中贴代码、禁止编造结果。回复不超过 2 句话。',
-    ].join('\n');
-
-    const msgs = await conversationService.buildContext(sessionId, conversationId, 'analyze', hint);
-    const sr = await conversationService.streamToSSE(msgs, writer, { temperature: 0.3, maxTokens: 150 });
-    await conversationService.finalizeAssistantMessage(placeholderMessageId, sr.content, sr.thinking, sr.tokens);
+    // 固定文本引导语
+    const hintText = `R 代码已生成（${generated.code.split('\n').length} 行），👉 请在右侧工作区核对代码并点击「确认并执行」。`;
+    await this.sendFixedHint(writer, placeholderMessageId, hintText);
 
     return { messageId: placeholderMessageId, intent: 'analyze', success: true };
   }
@@ -639,16 +654,11 @@ export class ChatHandlerService {
           durationMs,
         });
 
-        // 流式总结结果
-        const hint = [
-          `[系统指令] R 代码执行完成（${durationMs}ms），生成 ${(execResult.reportBlocks || []).length} 个报告模块。`,
-          '请简要解释结果的统计学意义。',
-          '【禁止】不要编造数值（工作区已展示完整结果）。',
-        ].join('\n');
-
-        const msgs = await conversationService.buildContext(sessionId, conversationId, 'analyze', hint);
-        const sr = await conversationService.streamToSSE(msgs, writer, { temperature: 0.5, maxTokens: 800 });
-        await conversationService.finalizeAssistantMessage(placeholderMessageId, sr.content, sr.thinking, sr.tokens);
+        // 固定文本引导语（结果解读应在右侧工作区，不在对话区）
+        const blockCount = (execResult.reportBlocks || []).length;
+        const seconds = (durationMs / 1000).toFixed(1);
+        const hintText = `✅ 分析完成（${seconds}s），共生成 ${blockCount} 个结果模块。👉 请在右侧工作区查看完整结果和图表。`;
+        await this.sendFixedHint(writer, placeholderMessageId, hintText);
 
         return { messageId: placeholderMessageId, intent: 'analyze', success: true };
       }