feat: 多功能升级 — 流式逐字渲染、对话缓存、会话组织优化、记忆管理修复、性能仪表盘

- 前端消息流式逐字渲染 (AI-Core ChatStream → SSE → Gateway → WebSocket stream_chunk → fadeInUp + cursorBlink)
- 后端对话缓存 (conversationCache sync.Map, GET /sessions/:id/messages)
- 前端侧边栏历史多轮对话显示
- DevTools 性能监控图标移至首页仪表盘
- DevTools 用户记忆查询/删减功能修复 (补全 DELETE 数据链路)
- 后端和 DevTools 按用户分类组织实时活动会话 (map[userID]map[sessionID]*Client)
- 新增 docs/api-reference/ 路由参考文档
- 新增 docs/message-flow-architecture.md 消息链路架构文档

devtools/src/index.js

@@ -198,6 +198,12 @@ app.get('/api/sessions', async (_req, res) => {
   res.status(result.status).json(result.body);
 });
 
+// GET /api/sessions/active — 获取按用户分组的活跃会话
+app.get('/api/sessions/active', async (_req, res) => {
+  const result = await proxyToGateway('/api/v1/sessions/active');
+  res.status(result.status).json(result.body);
+});
+
 app.get('/api/sessions/:id', async (req, res) => {
   const result = await proxyToGateway(`/api/v1/admin/sessions/${req.params.id}`);
   res.status(result.status).json(result.body);
@@ -230,6 +236,14 @@ app.post('/api/memory/add', async (req, res) => {
   res.status(result.status).json(result.body);
 });
 
+app.delete('/api/memory/:id', async (req, res) => {
+  const { id } = req.params;
+  if (!id) return res.status(400).json({ error: '缺少 memory id' });
+  const qs = new URLSearchParams({ id }).toString();
+  const result = await proxyToGateway(`/api/v1/memory?${qs}`, { method: 'DELETE' });
+  res.status(result.status).json(result.body);
+});
+
 // ---- 服务状态 ----
 app.get('/api/services', (_req, res) => {
   res.json(processManager.getStatus());
@@ -311,6 +325,16 @@ app.get('/api/performance', async (_req, res) => {
   res.json(snapshot);
 });
 
+// 性能仪表盘聚合数据 (供首页仪表盘使用)
+app.get('/api/performance/dashboard', async (_req, res) => {
+  try {
+    const dashboardData = await performanceMonitor.updateDashboard();
+    res.json(dashboardData);
+  } catch (err) {
+    res.status(500).json({ error: `获取性能仪表盘数据失败: ${err.message}` });
+  }
+});
+
 app.get('/api/performance/history', (_req, res) => {
   res.json(performanceMonitor.getAllHistory());
 });