fix(网关): 添加 model 参数必填验证

在以下端点添加 model 参数的必填验证，缺失时直接返回 400 错误：
- /v1/messages
- /v1/messages/count_tokens
- /openai/v1/responses

修复前：空 model 会进入账号选择流程，最终由上游 API 返回错误
修复后：入口处直接拒绝，避免浪费资源和不明确的错误信息

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

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

backend/internal/handler/gateway_handler.go

@@ -88,6 +88,12 @@ func (h *GatewayHandler) Messages(c *gin.Context) {
 	reqModel := parsedReq.Model
 	reqStream := parsedReq.Stream
 
+	// 验证 model 必填
+	if reqModel == "" {
+		h.errorResponse(c, http.StatusBadRequest, "invalid_request_error", "model is required")
+		return
+	}
+
 	// Track if we've started streaming (for error handling)
 	streamStarted := false
 
@@ -517,6 +523,12 @@ func (h *GatewayHandler) CountTokens(c *gin.Context) {
 		return
 	}
 
+	// 验证 model 必填
+	if parsedReq.Model == "" {
+		h.errorResponse(c, http.StatusBadRequest, "invalid_request_error", "model is required")
+		return
+	}
+
 	// 获取订阅信息（可能为nil）
 	subscription, _ := middleware2.GetSubscriptionFromContext(c)
 

backend/internal/handler/openai_gateway_handler.go

@@ -80,6 +80,12 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 	reqModel, _ := reqBody["model"].(string)
 	reqStream, _ := reqBody["stream"].(bool)
 
+	// 验证 model 必填
+	if reqModel == "" {
+		h.errorResponse(c, http.StatusBadRequest, "invalid_request_error", "model is required")
+		return
+	}
+
 	// For non-Codex CLI requests, set default instructions
 	userAgent := c.GetHeader("User-Agent")
 	if !openai.IsCodexCLIRequest(userAgent) {