Merge pull request #1110 from wangr0031/fix_parallel_tool_calls

feat: chat/completion路由透传parallel_tool_calls参数

controller/channel.go

@@ -119,8 +119,11 @@ func FetchUpstreamModels(c *gin.Context) {
 		baseURL = channel.GetBaseURL()
 	}
 	url := fmt.Sprintf("%s/v1/models", baseURL)
-	if channel.Type == common.ChannelTypeGemini {
+	switch channel.Type {
+	case common.ChannelTypeGemini:
 		url = fmt.Sprintf("%s/v1beta/openai/models", baseURL)
+	case common.ChannelTypeAli:
+		url = fmt.Sprintf("%s/compatible-mode/v1/models", baseURL)
 	}
 	body, err := GetResponseBody("GET", url, channel, GetAuthHeader(channel.Key))
 	if err != nil {

dto/openai_request.go

@@ -43,6 +43,7 @@ type GeneralOpenAIRequest struct {
 	ResponseFormat   *ResponseFormat   `json:"response_format,omitempty"`
 	EncodingFormat   any               `json:"encoding_format,omitempty"`
 	Seed             float64           `json:"seed,omitempty"`
+	ParallelTooCalls bool              `json:"parallel_tool_calls,omitempty"`
 	Tools            []ToolCallRequest `json:"tools,omitempty"`
 	ToolChoice       any               `json:"tool_choice,omitempty"`
 	User             string            `json:"user,omitempty"`
@@ -53,6 +54,7 @@ type GeneralOpenAIRequest struct {
 	Audio            any               `json:"audio,omitempty"`
 	EnableThinking   any               `json:"enable_thinking,omitempty"` // ali
 	ExtraBody        any               `json:"extra_body,omitempty"`
+	WebSearchOptions *WebSearchOptions `json:"web_search_options,omitempty"`
 }
 
 type ToolCallRequest struct {
@@ -371,6 +373,11 @@ func (m *Message) ParseContent() []MediaContent {
 	return contentList
 }
 
+type WebSearchOptions struct {
+	SearchContextSize string          `json:"search_context_size,omitempty"`
+	UserLocation      json.RawMessage `json:"user_location,omitempty"`
+}
+
 type OpenAIResponsesRequest struct {
 	Model              string               `json:"model"`
 	Input              json.RawMessage      `json:"input,omitempty"`

relay/channel/ali/adaptor.go

@@ -57,6 +57,12 @@ func (a *Adaptor) ConvertOpenAIRequest(c *gin.Context, info *relaycommon.RelayIn
 	if request == nil {
 		return nil, errors.New("request is nil")
 	}
+
+	// fix: ali parameter.enable_thinking must be set to false for non-streaming calls
+	if !info.IsStream {
+		request.EnableThinking = false
+	}
+
 	switch info.RelayMode {
 	default:
 		aliReq := requestOpenAI2Ali(*request)

relay/channel/vertex/adaptor.go

@@ -95,14 +95,23 @@ func (a *Adaptor) GetRequestURL(info *relaycommon.RelayInfo) (string, error) {
 		} else {
 			suffix = "generateContent"
 		}
-		return fmt.Sprintf(
-			"https://%s-aiplatform.googleapis.com/v1/projects/%s/locations/%s/publishers/google/models/%s:%s",
-			region,
-			adc.ProjectID,
-			region,
-			info.UpstreamModelName,
-			suffix,
-		), nil
+		if region == "global" {
+			return fmt.Sprintf(
+				"https://aiplatform.googleapis.com/v1/projects/%s/locations/global/publishers/google/models/%s:%s",
+				adc.ProjectID,
+				info.UpstreamModelName,
+				suffix,
+			), nil
+		} else {
+			return fmt.Sprintf(
+				"https://%s-aiplatform.googleapis.com/v1/projects/%s/locations/%s/publishers/google/models/%s:%s",
+				region,
+				adc.ProjectID,
+				region,
+				info.UpstreamModelName,
+				suffix,
+			), nil
+		}
 	} else if a.RequestMode == RequestModeClaude {
 		if info.IsStream {
 			suffix = "streamRawPredict?alt=sse"

relay/relay-text.go

@@ -47,6 +47,20 @@ func getAndValidateTextRequest(c *gin.Context, relayInfo *relaycommon.RelayInfo)
 	if textRequest.Model == "" {
 		return nil, errors.New("model is required")
 	}
+	if textRequest.WebSearchOptions != nil {
+		if textRequest.WebSearchOptions.SearchContextSize != "" {
+			validSizes := map[string]bool{
+				"high":   true,
+				"medium": true,
+				"low":    true,
+			}
+			if !validSizes[textRequest.WebSearchOptions.SearchContextSize] {
+				return nil, errors.New("invalid search_context_size, must be one of: high, medium, low")
+			}
+		} else {
+			textRequest.WebSearchOptions.SearchContextSize = "medium"
+		}
+	}
 	switch relayInfo.RelayMode {
 	case relayconstant.RelayModeCompletions:
 		if textRequest.Prompt == "" {
@@ -76,6 +90,10 @@ func TextHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) {
 
 	// get & validate textRequest 获取并验证文本请求
 	textRequest, err := getAndValidateTextRequest(c, relayInfo)
+	if textRequest.WebSearchOptions != nil {
+		c.Set("chat_completion_web_search_context_size", textRequest.WebSearchOptions.SearchContextSize)
+	}
+
 	if err != nil {
 		common.LogError(c, fmt.Sprintf("getAndValidateTextRequest failed: %s", err.Error()))
 		return service.OpenAIErrorWrapperLocal(err, "invalid_text_request", http.StatusBadRequest)
@@ -370,9 +388,20 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 			dWebSearchQuota = decimal.NewFromFloat(webSearchPrice).
 				Mul(decimal.NewFromInt(int64(webSearchTool.CallCount))).
 				Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
-			extraContent += fmt.Sprintf("Web Search 调用 %d 次，上下文大小 %s，调用花费 $%s",
+			extraContent += fmt.Sprintf("Web Search 调用 %d 次，上下文大小 %s，调用花费 %s",
 				webSearchTool.CallCount, webSearchTool.SearchContextSize, dWebSearchQuota.String())
 		}
+	} else if strings.HasSuffix(modelName, "search-preview") {
+		// search-preview 模型不支持 response api
+		searchContextSize := ctx.GetString("chat_completion_web_search_context_size")
+		if searchContextSize == "" {
+			searchContextSize = "medium"
+		}
+		webSearchPrice = operation_setting.GetWebSearchPricePerThousand(modelName, searchContextSize)
+		dWebSearchQuota = decimal.NewFromFloat(webSearchPrice).
+			Div(decimal.NewFromInt(1000)).Mul(dGroupRatio).Mul(dQuotaPerUnit)
+		extraContent += fmt.Sprintf("Web Search 调用 1 次，上下文大小 %s，调用花费 %s",
+			searchContextSize, dWebSearchQuota.String())
 	}
 	// file search tool 计费
 	var dFileSearchQuota decimal.Decimal
@@ -463,10 +492,16 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 		other["image_ratio"] = imageRatio
 		other["image_output"] = imageTokens
 	}
-	if !dWebSearchQuota.IsZero() && relayInfo.ResponsesUsageInfo != nil {
-		if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists {
+	if !dWebSearchQuota.IsZero() {
+		if relayInfo.ResponsesUsageInfo != nil {
+			if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists {
+				other["web_search"] = true
+				other["web_search_call_count"] = webSearchTool.CallCount
+				other["web_search_price"] = webSearchPrice
+			}
+		} else if strings.HasSuffix(modelName, "search-preview") {
 			other["web_search"] = true
-			other["web_search_call_count"] = webSearchTool.CallCount
+			other["web_search_call_count"] = 1
 			other["web_search_price"] = webSearchPrice
 		}
 	}

web/src/components/ChannelsTable.js

@@ -888,9 +888,13 @@ const ChannelsTable = () => {
     const localIdSort = localStorage.getItem('id-sort') === 'true';
     const localPageSize =
       parseInt(localStorage.getItem('page-size')) || ITEMS_PER_PAGE;
+    const localEnableTagMode = localStorage.getItem('enable-tag-mode') === 'true';
+    const localEnableBatchDelete = localStorage.getItem('enable-batch-delete') === 'true';
     setIdSort(localIdSort);
     setPageSize(localPageSize);
-    loadChannels(0, localPageSize, localIdSort, enableTagMode)
+    setEnableTagMode(localEnableTagMode);
+    setEnableBatchDelete(localEnableBatchDelete);
+    loadChannels(0, localPageSize, localIdSort, localEnableTagMode)
       .then()
       .catch((reason) => {
         showError(reason);
@@ -1486,10 +1490,12 @@ const ChannelsTable = () => {
             {t('开启批量操作')}
           </Typography.Text>
           <Switch
+            checked={enableBatchDelete}
             label={t('开启批量操作')}
             uncheckedText={t('关')}
             aria-label={t('是否开启批量操作')}
             onChange={(v) => {
+              localStorage.setItem('enable-batch-delete', v + '');
               setEnableBatchDelete(v);
             }}
           />
@@ -1553,6 +1559,7 @@ const ChannelsTable = () => {
             uncheckedText={t('关')}
             aria-label={t('是否启用标签聚合')}
             onChange={(v) => {
+              localStorage.setItem('enable-tag-mode', v + '');
               setEnableTagMode(v);
               loadChannels(0, pageSize, idSort, v);
             }}

web/src/pages/Home/index.js

@@ -158,7 +158,7 @@ const Home = () => {
                   </p>
                   <p>
                     {t('OIDC 身份验证')}：
-                    {statusState?.status?.oidc === true
+                    {statusState?.status?.oidc_enabled === true
                       ? t('已启用')
                       : t('未启用')}
                   </p>

web/src/pages/Playground/Playground.js

@@ -64,8 +64,9 @@ const Playground = () => {
     },
   ];
 
+  const defaultModel = 'gpt-4o-mini';
   const [inputs, setInputs] = useState({
-    model: 'gpt-4o-mini',
+    model: defaultModel,
     group: '',
     max_tokens: 0,
     temperature: 0,
@@ -108,6 +109,11 @@ const Playground = () => {
         value: model,
       }));
       setModels(localModelOptions);
+      // if default model is not in the list, set the first one as default
+      const hasDefault = localModelOptions.some(option => option.value === defaultModel);
+      if (!hasDefault && localModelOptions.length > 0) {
+        setInputs((inputs) => ({ ...inputs, model: localModelOptions[0].value }));
+      }
     } else {
       showError(t(message));
     }