fix(流式): 提升SSE稳定性并统一超时配置

- 扩展SSE行长与间隔超时处理,补充keepalive

- 写入失败与超长行时发送错误事件,修复并发释放

- 同步默认配置与示例配置,更新Caddy超时/压缩规则

- 新增OpenAI流式超时与超长行测试

测试: go test ./...
This commit is contained in:
yangjianbo
2026-01-04 19:49:59 +08:00
parent 6c86cf7605
commit 73ffb58518
14 changed files with 721 additions and 146 deletions

View File

@@ -165,7 +165,7 @@ func (h *GatewayHandler) GeminiV1BetaModels(c *gin.Context) {
subscription, _ := middleware.GetSubscriptionFromContext(c)
// For Gemini native API, do not send Claude-style ping frames.
geminiConcurrency := NewConcurrencyHelper(h.concurrencyHelper.concurrencyService, SSEPingFormatNone)
geminiConcurrency := NewConcurrencyHelper(h.concurrencyHelper.concurrencyService, SSEPingFormatNone, 0)
// 0) wait queue check
maxWait := service.CalculateMaxWait(authSubject.Concurrency)
@@ -185,6 +185,8 @@ func (h *GatewayHandler) GeminiV1BetaModels(c *gin.Context) {
googleError(c, http.StatusTooManyRequests, err.Error())
return
}
// 确保请求取消时也会释放槽位,避免长连接被动中断造成泄漏
userReleaseFunc = wrapReleaseOnDone(c.Request.Context(), userReleaseFunc)
if userReleaseFunc != nil {
defer userReleaseFunc()
}
@@ -261,6 +263,9 @@ func (h *GatewayHandler) GeminiV1BetaModels(c *gin.Context) {
log.Printf("Bind sticky session failed: %v", err)
}
}
// 账号槽位/等待计数需要在超时或断开时安全回收
accountReleaseFunc = wrapReleaseOnDone(c.Request.Context(), accountReleaseFunc)
accountWaitRelease = wrapReleaseOnDone(c.Request.Context(), accountWaitRelease)
// 5) forward (根据平台分流)
var result *service.ForwardResult