fix: 优化 API Key 并发控制机制

- 调整并发计数器过期时间为3分钟，支持长时间流式请求
- 为流式响应添加客户端断开检测，确保计数正确减少
- 添加响应关闭和错误事件监听器，防止并发计数泄漏
- 提高系统稳定性和资源管理准确性

🤖 Generated with [Claude Code](https://claude.ai/code)

Co-Authored-By: Claude <noreply@anthropic.com>

src/models/redis.js

@@ -721,8 +721,10 @@ class RedisClient {
       const key = `concurrency:${apiKeyId}`;
       const count = await this.client.incr(key);
       
-      // 设置过期时间为5分钟，防止计数器永远不清零
-      await this.client.expire(key, 300);
+      // 设置过期时间为180秒（3分钟），防止计数器永远不清零
+      // 正常情况下请求会在完成时主动减少计数，这只是一个安全保障
+      // 180秒足够支持较长的流式请求
+      await this.client.expire(key, 180);
       
       logger.database(`🔢 Incremented concurrency for key ${apiKeyId}: ${count}`);
       return count;

src/routes/api.js

@@ -45,6 +45,21 @@ router.post('/v1/messages', authenticateApiKey, async (req, res) => {
       res.setHeader('Connection', 'keep-alive');
       res.setHeader('Access-Control-Allow-Origin', '*');
       
+      // 为流式响应添加客户端断开检测，确保并发计数正确减少
+      if (req.concurrencyInfo) {
+        // 添加响应关闭事件监听器
+        res.on('close', () => {
+          logger.api(`🔌 Stream response closed for key: ${req.apiKey.id} (${req.apiKey.name}), triggering concurrency decrement`);
+          req.concurrencyInfo.decrementConcurrency();
+        });
+        
+        // 添加错误事件监听器
+        res.on('error', (error) => {
+          logger.api(`⚠️ Stream response error for key: ${req.apiKey.id} (${req.apiKey.name}): ${error.message}`);
+          req.concurrencyInfo.decrementConcurrency();
+        });
+      }
+      
       let usageDataCaptured = false;
       
       // 使用自定义流处理器来捕获usage数据