Merge upstream/main

backend/internal/handler/openai_gateway_handler.go

@@ -8,9 +8,11 @@ import (
 	"io"
 	"log"
 	"net/http"
+	"strings"
 	"time"
 
 	"github.com/Wei-Shaw/sub2api/internal/config"
+	"github.com/Wei-Shaw/sub2api/internal/pkg/ip"
 	"github.com/Wei-Shaw/sub2api/internal/pkg/openai"
 	middleware2 "github.com/Wei-Shaw/sub2api/internal/server/middleware"
 	"github.com/Wei-Shaw/sub2api/internal/service"
@@ -81,6 +83,8 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 		return
 	}
 
+	setOpsRequestContext(c, "", false, body)
+
 	// Parse request body to map for potential modification
 	var reqBody map[string]any
 	if err := json.Unmarshal(body, &reqBody); err != nil {
@@ -98,15 +102,41 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 		return
 	}
 
-	// For non-Codex CLI requests, set default instructions
 	userAgent := c.GetHeader("User-Agent")
 	if !openai.IsCodexCLIRequest(userAgent) {
-		reqBody["instructions"] = openai.DefaultInstructions
-		// Re-serialize body
-		body, err = json.Marshal(reqBody)
-		if err != nil {
-			h.errorResponse(c, http.StatusInternalServerError, "api_error", "Failed to process request")
-			return
+		existingInstructions, _ := reqBody["instructions"].(string)
+		if strings.TrimSpace(existingInstructions) == "" {
+			if instructions := strings.TrimSpace(service.GetOpenCodeInstructions()); instructions != "" {
+				reqBody["instructions"] = instructions
+				// Re-serialize body
+				body, err = json.Marshal(reqBody)
+				if err != nil {
+					h.errorResponse(c, http.StatusInternalServerError, "api_error", "Failed to process request")
+					return
+				}
+			}
+		}
+	}
+
+	setOpsRequestContext(c, reqModel, reqStream, body)
+
+	// 提前校验 function_call_output 是否具备可关联上下文，避免上游 400。
+	// 要求 previous_response_id，或 input 内存在带 call_id 的 tool_call/function_call，
+	// 或带 id 且与 call_id 匹配的 item_reference。
+	if service.HasFunctionCallOutput(reqBody) {
+		previousResponseID, _ := reqBody["previous_response_id"].(string)
+		if strings.TrimSpace(previousResponseID) == "" && !service.HasToolCallContext(reqBody) {
+			if service.HasFunctionCallOutputMissingCallID(reqBody) {
+				log.Printf("[OpenAI Handler] function_call_output 缺少 call_id: model=%s", reqModel)
+				h.errorResponse(c, http.StatusBadRequest, "invalid_request_error", "function_call_output requires call_id or previous_response_id; if relying on history, ensure store=true and reuse previous_response_id")
+				return
+			}
+			callIDs := service.FunctionCallOutputCallIDs(reqBody)
+			if !service.HasItemReferenceForCallIDs(reqBody, callIDs) {
+				log.Printf("[OpenAI Handler] function_call_output 缺少匹配的 item_reference: model=%s", reqModel)
+				h.errorResponse(c, http.StatusBadRequest, "invalid_request_error", "function_call_output requires item_reference ids matching each call_id, or previous_response_id/tool_call context; if relying on history, ensure store=true and reuse previous_response_id")
+				return
+			}
 		}
 	}
 
@@ -119,6 +149,7 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 	// 0. Check if wait queue is full
 	maxWait := service.CalculateMaxWait(subject.Concurrency)
 	canWait, err := h.concurrencyHelper.IncrementWaitCount(c.Request.Context(), subject.UserID, maxWait)
+	waitCounted := false
 	if err != nil {
 		log.Printf("Increment wait count failed: %v", err)
 		// On error, allow request to proceed
@@ -126,8 +157,14 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 		h.errorResponse(c, http.StatusTooManyRequests, "rate_limit_error", "Too many pending requests, please retry later")
 		return
 	}
-	// Ensure wait count is decremented when function exits
-	defer h.concurrencyHelper.DecrementWaitCount(c.Request.Context(), subject.UserID)
+	if err == nil && canWait {
+		waitCounted = true
+	}
+	defer func() {
+		if waitCounted {
+			h.concurrencyHelper.DecrementWaitCount(c.Request.Context(), subject.UserID)
+		}
+	}()
 
 	// 1. First acquire user concurrency slot
 	userReleaseFunc, err := h.concurrencyHelper.AcquireUserSlotWithWait(c, subject.UserID, subject.Concurrency, reqStream, &streamStarted)
@@ -136,6 +173,11 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 		h.handleConcurrencyError(c, err, "user", streamStarted)
 		return
 	}
+	// User slot acquired: no longer waiting.
+	if waitCounted {
+		h.concurrencyHelper.DecrementWaitCount(c.Request.Context(), subject.UserID)
+		waitCounted = false
+	}
 	// 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 {
@@ -173,15 +215,16 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 		}
 		account := selection.Account
 		log.Printf("[OpenAI Handler] Selected account: id=%d name=%s", account.ID, account.Name)
+		setOpsSelectedAccount(c, account.ID)
 
 		// 3. Acquire account concurrency slot
 		accountReleaseFunc := selection.ReleaseFunc
-		var accountWaitRelease func()
 		if !selection.Acquired {
 			if selection.WaitPlan == nil {
 				h.handleStreamingAwareError(c, http.StatusServiceUnavailable, "api_error", "No available accounts", streamStarted)
 				return
 			}
+			accountWaitCounted := false
 			canWait, err := h.concurrencyHelper.IncrementAccountWaitCount(c.Request.Context(), account.ID, selection.WaitPlan.MaxWaiting)
 			if err != nil {
 				log.Printf("Increment account wait count failed: %v", err)
@@ -189,12 +232,15 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 				log.Printf("Account wait queue full: account=%d", account.ID)
 				h.handleStreamingAwareError(c, http.StatusTooManyRequests, "rate_limit_error", "Too many pending requests, please retry later", streamStarted)
 				return
-			} else {
-				// Only set release function if increment succeeded
-				accountWaitRelease = func() {
+			}
+			if err == nil && canWait {
+				accountWaitCounted = true
+			}
+			defer func() {
+				if accountWaitCounted {
 					h.concurrencyHelper.DecrementAccountWaitCount(c.Request.Context(), account.ID)
 				}
-			}
+			}()
 
 			accountReleaseFunc, err = h.concurrencyHelper.AcquireAccountSlotWithWaitTimeout(
 				c,
@@ -205,29 +251,26 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 				&streamStarted,
 			)
 			if err != nil {
-				if accountWaitRelease != nil {
-					accountWaitRelease()
-				}
 				log.Printf("Account concurrency acquire failed: %v", err)
 				h.handleConcurrencyError(c, err, "account", streamStarted)
 				return
 			}
+			if accountWaitCounted {
+				h.concurrencyHelper.DecrementAccountWaitCount(c.Request.Context(), account.ID)
+				accountWaitCounted = false
+			}
 			if err := h.gatewayService.BindStickySession(c.Request.Context(), apiKey.GroupID, sessionHash, account.ID); err != nil {
 				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)
 
 		// Forward request
 		result, err := h.gatewayService.Forward(c.Request.Context(), c, account, body)
 		if accountReleaseFunc != nil {
 			accountReleaseFunc()
 		}
-		if accountWaitRelease != nil {
-			accountWaitRelease()
-		}
 		if err != nil {
 			var failoverErr *service.UpstreamFailoverError
 			if errors.As(err, &failoverErr) {
@@ -247,8 +290,12 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 			return
 		}
 
+		// 捕获请求信息（用于异步记录，避免在 goroutine 中访问 gin.Context）
+		userAgent := c.GetHeader("User-Agent")
+		clientIP := ip.GetClientIP(c)
+
 		// Async record usage
-		go func(result *service.OpenAIForwardResult, usedAccount *service.Account, ua string) {
+		go func(result *service.OpenAIForwardResult, usedAccount *service.Account, ua, ip string) {
 			ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 			defer cancel()
 			if err := h.gatewayService.RecordUsage(ctx, &service.OpenAIRecordUsageInput{
@@ -258,10 +305,11 @@ func (h *OpenAIGatewayHandler) Responses(c *gin.Context) {
 				Account:      usedAccount,
 				Subscription: subscription,
 				UserAgent:    ua,
+				IPAddress:    ip,
 			}); err != nil {
 				log.Printf("Record usage failed: %v", err)
 			}
-		}(result, account, userAgent)
+		}(result, account, userAgent, clientIP)
 		return
 	}
 }