refactor(responses): remove compact model suffix handling (#6770)

common/api_type.go

@@ -88,7 +88,7 @@ func ChannelType2APIType(channelType int) (int, bool) {
 	return apiType, true
 }
 
-func IsResponsesCompactAPIType(apiType int) bool {
+func SupportsResponsesCompact(channelType, apiType int) bool {
 	switch apiType {
 	case constant.APITypeOpenAI,
 		constant.APITypeCodex,

controller/channel-test.go

@@ -27,7 +27,6 @@ import (
 	"github.com/QuantumNous/new-api/relaykit/types"
 	"github.com/QuantumNous/new-api/service"
 	"github.com/QuantumNous/new-api/setting/operation_setting"
-	"github.com/QuantumNous/new-api/setting/ratio_setting"
 	hosttypes "github.com/QuantumNous/new-api/types"
 
 	"github.com/samber/lo"
@@ -42,14 +41,11 @@ type testResult struct {
 	newAPIError *types.NewAPIError
 }
 
-func normalizeChannelTestEndpoint(channel *model.Channel, modelName, endpointType string) string {
+func normalizeChannelTestEndpoint(channel *model.Channel, endpointType string) string {
 	normalized := strings.TrimSpace(endpointType)
 	if normalized != "" {
 		return normalized
 	}
-	if strings.HasSuffix(modelName, ratio_setting.CompactModelSuffix) {
-		return string(constant.EndpointTypeOpenAIResponseCompact)
-	}
 	if channel != nil && channel.Type == constant.ChannelTypeCodex {
 		return string(constant.EndpointTypeOpenAIResponse)
 	}
@@ -111,7 +107,7 @@ func testChannel(ctx context.Context, channel *model.Channel, testUserID int, te
 		}
 	}
 
-	endpointType = normalizeChannelTestEndpoint(channel, testModel, endpointType)
+	endpointType = normalizeChannelTestEndpoint(channel, endpointType)
 
 	requestPath := "/v1/chat/completions"
 
@@ -146,20 +142,12 @@ func testChannel(ctx context.Context, channel *model.Channel, testUserID int, te
 			requestPath = "/v1/responses"
 		}
 
-		// responses compaction models (must use /v1/responses/compact)
-		if strings.HasSuffix(testModel, ratio_setting.CompactModelSuffix) {
-			requestPath = "/v1/responses/compact"
-		}
 	}
 	// Gemini 原生流式通过 URL action（:streamGenerateContent）表达而非请求体字段，
 	// GeminiChatRequest.IsStream 依据请求 URL 判定，合成请求路径需与生产入口保持一致
 	if isStream && constant.EndpointType(endpointType) == constant.EndpointTypeGemini {
 		requestPath = strings.Replace(requestPath, ":generateContent", ":streamGenerateContent", 1)
 	}
-	if strings.HasPrefix(requestPath, "/v1/responses/compact") {
-		testModel = ratio_setting.WithCompactModelSuffix(testModel)
-	}
-
 	c.Request = httptest.NewRequestWithContext(ctx, http.MethodPost, requestPath, nil)
 
 	cache, err := model.GetUserCache(testUserID)
@@ -277,7 +265,7 @@ func testChannel(ctx context.Context, channel *model.Channel, testUserID int, te
 
 	apiType, _ := common.ChannelType2APIType(channel.Type)
 	if info.RelayMode == relayconstant.RelayModeResponsesCompact &&
-		!common.IsResponsesCompactAPIType(apiType) {
+		!common.SupportsResponsesCompact(channel.Type, apiType) {
 		return testResult{
 			context:     c,
 			localErr:    fmt.Errorf("responses compaction test is not supported for api type %d", apiType),
@@ -806,14 +794,6 @@ func buildTestRequest(model string, endpointType string, channel *model.Channel,
 		}
 	}
 
-	// Responses compaction models (must use /v1/responses/compact)
-	if strings.HasSuffix(model, ratio_setting.CompactModelSuffix) {
-		return &dto.OpenAIResponsesCompactionRequest{
-			Model: model,
-			Input: testResponsesInput,
-		}
-	}
-
 	// Responses-only models (e.g. codex series)
 	if strings.Contains(strings.ToLower(model), "codex") {
 		return &dto.OpenAIResponsesRequest{

controller/channel_test_internal_test.go

@@ -95,23 +95,25 @@ func TestNewAPIChannelRegistration(t *testing.T) {
 	assert.Empty(t, constant.ChannelBaseURLs[constant.ChannelTypeNewAPI])
 }
 
-func TestResponsesCompactAPITypeSupport(t *testing.T) {
+func TestResponsesCompactChannelSupport(t *testing.T) {
 	tests := []struct {
-		name    string
-		apiType int
-		want    bool
+		name        string
+		channelType int
+		apiType     int
+		want        bool
 	}{
-		{name: "OpenAI", apiType: constant.APITypeOpenAI, want: true},
-		{name: "Codex", apiType: constant.APITypeCodex, want: true},
-		{name: "Advanced Custom", apiType: constant.APITypeAdvancedCustom, want: true},
-		{name: "Sub2API", apiType: constant.APITypeSub2API, want: true},
-		{name: "New API", apiType: constant.APITypeNewAPI, want: true},
-		{name: "Anthropic", apiType: constant.APITypeAnthropic, want: false},
+		{name: "OpenAI", channelType: constant.ChannelTypeOpenAI, apiType: constant.APITypeOpenAI, want: true},
+		{name: "Azure", channelType: constant.ChannelTypeAzure, apiType: constant.APITypeOpenAI, want: true},
+		{name: "Codex", channelType: constant.ChannelTypeCodex, apiType: constant.APITypeCodex, want: true},
+		{name: "Advanced Custom", channelType: constant.ChannelTypeAdvancedCustom, apiType: constant.APITypeAdvancedCustom, want: true},
+		{name: "Sub2API", channelType: constant.ChannelTypeSub2API, apiType: constant.APITypeSub2API, want: true},
+		{name: "New API", channelType: constant.ChannelTypeNewAPI, apiType: constant.APITypeNewAPI, want: true},
+		{name: "Anthropic", channelType: constant.ChannelTypeAnthropic, apiType: constant.APITypeAnthropic, want: false},
 	}
 
 	for _, test := range tests {
 		t.Run(test.name, func(t *testing.T) {
-			assert.Equal(t, test.want, common.IsResponsesCompactAPIType(test.apiType))
+			assert.Equal(t, test.want, common.SupportsResponsesCompact(test.channelType, test.apiType))
 		})
 	}
 }

middleware/distributor.go

@@ -410,9 +410,6 @@ func getModelRequest(c *gin.Context) (*ModelRequest, bool, error) {
 		common.SetContextKey(c, constant.ContextKeyTokenGroup, modelRequest.Group)
 	}
 
-	if strings.HasPrefix(c.Request.URL.Path, "/v1/responses/compact") && modelRequest.Model != "" {
-		modelRequest.Model = ratio_setting.WithCompactModelSuffix(modelRequest.Model)
-	}
 	return &modelRequest, shouldSelectChannel, nil
 }
 

relay/channel/codex/constants.go

 package codex
 
-import (
-	"slices"
-
-	"github.com/QuantumNous/new-api/setting/ratio_setting"
-)
-
-var baseModelList = []string{
+var ModelList = []string{
 	"gpt-5.6-sol",
 	"gpt-5.6-terra",
 	"gpt-5.6-luna",
@@ -17,11 +11,4 @@ var baseModelList = []string{
 	"codex-auto-review",
 }
 
-var ModelList = slices.DeleteFunc(
-	ratio_setting.WithCompactModelVariants(baseModelList),
-	func(modelName string) bool {
-		return modelName == ratio_setting.WithCompactModelSuffix("codex-auto-review")
-	},
-)
-
 const ChannelName = "codex"

relay/helper/model_mapped.go

@@ -4,12 +4,9 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
-	"strings"
 
 	"github.com/QuantumNous/new-api/relay/common"
-	relayconstant "github.com/QuantumNous/new-api/relay/constant"
 	"github.com/QuantumNous/new-api/relaykit/dto"
-	"github.com/QuantumNous/new-api/setting/ratio_setting"
 	"github.com/gin-gonic/gin"
 )
 
@@ -18,13 +15,6 @@ func ModelMappedHelper(c *gin.Context, info *common.RelayInfo, request dto.Reque
 		info.ChannelMeta = &common.ChannelMeta{}
 	}
 
-	isResponsesCompact := info.RelayMode == relayconstant.RelayModeResponsesCompact
-	originModelName := info.OriginModelName
-	mappingModelName := originModelName
-	if isResponsesCompact && strings.HasSuffix(originModelName, ratio_setting.CompactModelSuffix) {
-		mappingModelName = strings.TrimSuffix(originModelName, ratio_setting.CompactModelSuffix)
-	}
-
 	// map model name
 	modelMapping := c.GetString("model_mapping")
 	if modelMapping != "" && modelMapping != "{}" {
@@ -35,7 +25,7 @@ func ModelMappedHelper(c *gin.Context, info *common.RelayInfo, request dto.Reque
 		}
 
 		// 支持链式模型重定向，最终使用链尾的模型
-		currentModel := mappingModelName
+		currentModel := info.OriginModelName
 		visitedModels := map[string]bool{
 			currentModel: true,
 		}
@@ -66,14 +56,6 @@ func ModelMappedHelper(c *gin.Context, info *common.RelayInfo, request dto.Reque
 		}
 	}
 
-	if isResponsesCompact {
-		finalUpstreamModelName := mappingModelName
-		if info.IsModelMapped && info.UpstreamModelName != "" {
-			finalUpstreamModelName = info.UpstreamModelName
-		}
-		info.UpstreamModelName = finalUpstreamModelName
-		info.OriginModelName = ratio_setting.WithCompactModelSuffix(finalUpstreamModelName)
-	}
 	if request != nil {
 		request.SetModelName(info.UpstreamModelName)
 	}

relay/responses_handler.go

@@ -22,7 +22,7 @@ import (
 func ResponsesHelper(c *gin.Context, info *relaycommon.RelayInfo) (newAPIError *types.NewAPIError) {
 	info.InitChannelMeta(c)
 	if info.RelayMode == relayconstant.RelayModeResponsesCompact &&
-		!common.IsResponsesCompactAPIType(info.ApiType) {
+		!common.SupportsResponsesCompact(info.ChannelType, info.ApiType) {
 		return types.NewErrorWithStatusCode(
 			fmt.Errorf("unsupported endpoint %q for api type %d", "/v1/responses/compact", info.ApiType),
 			types.ErrorCodeInvalidRequest,

service/codex_channel_models.go

@@ -9,7 +9,6 @@ import (
 
 	"github.com/QuantumNous/new-api/constant"
 	"github.com/QuantumNous/new-api/model"
-	"github.com/QuantumNous/new-api/setting/ratio_setting"
 )
 
 func FetchCodexChannelModels(channel *model.Channel) ([]string, error) {
@@ -78,13 +77,5 @@ func fetchCodexChannelModels(
 	if statusCode < http.StatusOK || statusCode >= http.StatusMultipleChoices {
 		return nil, fmt.Errorf("upstream status: %d", statusCode)
 	}
-	modelVariants := make([]string, 0, len(models)*2)
-	modelVariants = append(modelVariants, models...)
-	for _, modelName := range models {
-		if modelName == "codex-auto-review" {
-			continue
-		}
-		modelVariants = append(modelVariants, ratio_setting.WithCompactModelSuffix(modelName))
-	}
-	return modelVariants, nil
+	return models, nil
 }

setting/ratio_setting/compact_suffix.go

-package ratio_setting
-
-import "strings"
-
-const CompactModelSuffix = "-openai-compact"
-const CompactWildcardModelKey = "*" + CompactModelSuffix
-
-func WithCompactModelSuffix(modelName string) string {
-	if strings.HasSuffix(modelName, CompactModelSuffix) {
-		return modelName
-	}
-	return modelName + CompactModelSuffix
-}
-
-func WithCompactModelVariants(models []string) []string {
-	variants := make([]string, 0, len(models)*2)
-	seen := make(map[string]struct{}, len(models)*2)
-	for _, model := range models {
-		if _, ok := seen[model]; ok {
-			continue
-		}
-		seen[model] = struct{}{}
-		variants = append(variants, model)
-	}
-	for _, model := range models {
-		compactModel := WithCompactModelSuffix(model)
-		if _, ok := seen[compactModel]; ok {
-			continue
-		}
-		seen[compactModel] = struct{}{}
-		variants = append(variants, compactModel)
-	}
-	return variants
-}

setting/ratio_setting/model_ratio.go

@@ -364,17 +364,6 @@ func GetModelPrice(name string, printErr bool) (float64, bool) {
 		return price, true
 	}
 
-	if strings.HasSuffix(name, CompactModelSuffix) {
-		price, ok := modelPriceMap.Get(CompactWildcardModelKey)
-		if !ok {
-			if printErr {
-				common.SysError("model price not found: " + name)
-			}
-			return -1, false
-		}
-		return price, true
-	}
-
 	if printErr {
 		common.SysError("model price not found: " + name)
 	}
@@ -398,12 +387,6 @@ func GetModelRatio(name string) (float64, bool, string) {
 
 	ratio, ok := modelRatioMap.Get(name)
 	if !ok {
-		if strings.HasSuffix(name, CompactModelSuffix) {
-			if wildcardRatio, ok := modelRatioMap.Get(CompactWildcardModelKey); ok {
-				return wildcardRatio, true, name
-			}
-			//return 0, true, name
-		}
 		return 37.5, operation_setting.SelfUseModeEnabled, name
 	}
 	return ratio, true, name