From 6a5f6b96b6d438bd785ec14bda8840ba80ff6c29 Mon Sep 17 00:00:00 2001
From: name <136912576+is7Qin@users.noreply.github.com>
Date: Sun, 31 May 2026 01:01:58 +0800
Subject: [PATCH] refactor(gateway): introduce OpenAI request view

Cache hot-path request scalars before full body decoding so later branches can avoid repeated map work while preserving current decode behavior.
---
 .../service/openai_gateway_service.go         | 45 ++++++++++++++-----
 .../openai_gateway_service_hotpath_test.go    | 20 +++++++++
 2 files changed, 55 insertions(+), 10 deletions(-)

diff --git a/backend/internal/service/openai_gateway_service.go b/backend/internal/service/openai_gateway_service.go
index d1460e5a08..87ffb2cb21 100644
--- a/backend/internal/service/openai_gateway_service.go
+++ b/backend/internal/service/openai_gateway_service.go
@@ -2346,7 +2346,8 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 	}
 
 	originalBody := body
-	reqModel, reqStream, promptCacheKey := extractOpenAIRequestMetaFromBody(body)
+	requestView := newOpenAIRequestView(body)
+	reqModel, reqStream, promptCacheKey := requestView.Model, requestView.Stream, requestView.PromptCacheKey
 	originalModel := reqModel
 
 	if account.Type == AccountTypeAPIKey && !openai_compat.ShouldUseResponsesAPI(account.Extra) {
@@ -2396,7 +2397,7 @@ func (s *OpenAIGatewayService) Forward(ctx context.Context, c *gin.Context, acco
 		return s.forwardOpenAIPassthrough(ctx, c, account, originalBody, reqModel, reasoningEffort, reqStream, startTime)
 	}
 
-	reqBody, err := getOpenAIRequestBodyMap(c, body)
+	reqBody, err := requestView.Decode(c)
 	if err != nil {
 		return nil, err
 	}
@@ -6274,15 +6275,39 @@ func deriveOpenAIReasoningEffortFromModel(model string) string {
 	return normalizeOpenAIReasoningEffort(parts[len(parts)-1])
 }
 
-func extractOpenAIRequestMetaFromBody(body []byte) (model string, stream bool, promptCacheKey string) {
-	if len(body) == 0 {
-		return "", false, ""
-	}
+type openAIRequestView struct {
+	body               []byte
+	Model              string
+	Stream             bool
+	PromptCacheKey     string
+	PreviousResponseID string
+	ServiceTier        string
+	ReasoningEffort    string
+}
 
-	model = strings.TrimSpace(gjson.GetBytes(body, "model").String())
-	stream = gjson.GetBytes(body, "stream").Bool()
-	promptCacheKey = strings.TrimSpace(gjson.GetBytes(body, "prompt_cache_key").String())
-	return model, stream, promptCacheKey
+func newOpenAIRequestView(body []byte) openAIRequestView {
+	if len(body) == 0 {
+		return openAIRequestView{}
+	}
+	return openAIRequestView{
+		body:               body,
+		Model:              strings.TrimSpace(gjson.GetBytes(body, "model").String()),
+		Stream:             gjson.GetBytes(body, "stream").Bool(),
+		PromptCacheKey:     strings.TrimSpace(gjson.GetBytes(body, "prompt_cache_key").String()),
+		PreviousResponseID: strings.TrimSpace(gjson.GetBytes(body, "previous_response_id").String()),
+		ServiceTier:        strings.TrimSpace(gjson.GetBytes(body, "service_tier").String()),
+		ReasoningEffort:    strings.TrimSpace(gjson.GetBytes(body, "reasoning.effort").String()),
+	}
+}
+
+// Decode 保留阶段一既有 full-map 行为；后续阶段会把调用点下沉到复杂分支。
+func (v openAIRequestView) Decode(c *gin.Context) (map[string]any, error) {
+	return getOpenAIRequestBodyMap(c, v.body)
+}
+
+func extractOpenAIRequestMetaFromBody(body []byte) (model string, stream bool, promptCacheKey string) {
+	view := newOpenAIRequestView(body)
+	return view.Model, view.Stream, view.PromptCacheKey
 }
 
 // normalizeOpenAIPassthroughOAuthBody 将透传 OAuth 请求体收敛为旧链路关键行为：
diff --git a/backend/internal/service/openai_gateway_service_hotpath_test.go b/backend/internal/service/openai_gateway_service_hotpath_test.go
index af0d21c4a1..df17ed2ef0 100644
--- a/backend/internal/service/openai_gateway_service_hotpath_test.go
+++ b/backend/internal/service/openai_gateway_service_hotpath_test.go
@@ -9,6 +9,26 @@ import (
 	"github.com/stretchr/testify/require"
 )
 
+func TestOpenAIRequestView_ExtractsRawScalars(t *testing.T) {
+	view := newOpenAIRequestView([]byte(`{"model":" gpt-5 ","stream":true,"prompt_cache_key":" ses-1 ","previous_response_id":" resp-1 ","service_tier":" fast ","reasoning":{"effort":" medium "}}`))
+
+	require.Equal(t, "gpt-5", view.Model)
+	require.True(t, view.Stream)
+	require.Equal(t, "ses-1", view.PromptCacheKey)
+	require.Equal(t, "resp-1", view.PreviousResponseID)
+	require.Equal(t, "fast", view.ServiceTier)
+	require.Equal(t, "medium", view.ReasoningEffort)
+}
+
+func TestOpenAIRequestView_DecodeKeepsFullMapBehavior(t *testing.T) {
+	view := newOpenAIRequestView([]byte(`{"model":"gpt-5","stream":true,"input":[{"type":"message","content":"hi"}]}`))
+
+	reqBody, err := view.Decode(nil)
+	require.NoError(t, err)
+	require.Equal(t, "gpt-5", reqBody["model"])
+	require.IsType(t, []any{}, reqBody["input"])
+}
+
 func TestExtractOpenAIRequestMetaFromBody(t *testing.T) {
 	tests := []struct {
 		name          string