Merge pull request #1025 from QuantumNous/responses_buildin_tools

feat: implement OpenAI responses built-in tool tracking
2025-05-07 02:25:47 +08:00
parent dc9e3b4139 c0095d4521
commit 65a15dbc17
12 changed files with 451 additions and 177 deletions
--- a/dto/openai_response.go
+++ b/dto/openai_response.go
@@ -195,28 +195,28 @@ type OutputTokenDetails struct {
 }

 type OpenAIResponsesResponse struct {
-	ID                 string             `json:"id"`
-	Object             string             `json:"object"`
-	CreatedAt          int                `json:"created_at"`
-	Status             string             `json:"status"`
-	Error              *OpenAIError       `json:"error,omitempty"`
-	IncompleteDetails  *IncompleteDetails `json:"incomplete_details,omitempty"`
-	Instructions       string             `json:"instructions"`
-	MaxOutputTokens    int                `json:"max_output_tokens"`
-	Model              string             `json:"model"`
-	Output             []ResponsesOutput  `json:"output"`
-	ParallelToolCalls  bool               `json:"parallel_tool_calls"`
-	PreviousResponseID string             `json:"previous_response_id"`
-	Reasoning          *Reasoning         `json:"reasoning"`
-	Store              bool               `json:"store"`
-	Temperature        float64            `json:"temperature"`
-	ToolChoice         string             `json:"tool_choice"`
-	Tools              []interface{}      `json:"tools"`
-	TopP               float64            `json:"top_p"`
-	Truncation         string             `json:"truncation"`
-	Usage              *Usage             `json:"usage"`
-	User               json.RawMessage    `json:"user"`
-	Metadata           json.RawMessage    `json:"metadata"`
+	ID                 string               `json:"id"`
+	Object             string               `json:"object"`
+	CreatedAt          int                  `json:"created_at"`
+	Status             string               `json:"status"`
+	Error              *OpenAIError         `json:"error,omitempty"`
+	IncompleteDetails  *IncompleteDetails   `json:"incomplete_details,omitempty"`
+	Instructions       string               `json:"instructions"`
+	MaxOutputTokens    int                  `json:"max_output_tokens"`
+	Model              string               `json:"model"`
+	Output             []ResponsesOutput    `json:"output"`
+	ParallelToolCalls  bool                 `json:"parallel_tool_calls"`
+	PreviousResponseID string               `json:"previous_response_id"`
+	Reasoning          *Reasoning           `json:"reasoning"`
+	Store              bool                 `json:"store"`
+	Temperature        float64              `json:"temperature"`
+	ToolChoice         string               `json:"tool_choice"`
+	Tools              []ResponsesToolsCall `json:"tools"`
+	TopP               float64              `json:"top_p"`
+	Truncation         string               `json:"truncation"`
+	Usage              *Usage               `json:"usage"`
+	User               json.RawMessage      `json:"user"`
+	Metadata           json.RawMessage      `json:"metadata"`
 }

 type IncompleteDetails struct {
@@ -238,8 +238,12 @@ type ResponsesOutputContent struct {
 }

 const (
-	BuildInTools_WebSearch  = "web_search_preview"
-	BuildInTools_FileSearch = "file_search"
+	BuildInToolWebSearchPreview = "web_search_preview"
+	BuildInToolFileSearch       = "file_search"
+)
+
+const (
+	BuildInCallWebSearchCall = "web_search_call"
 )

 const (
@@ -250,6 +254,7 @@ const (
 // ResponsesStreamResponse 用于处理 /v1/responses 流式响应
 type ResponsesStreamResponse struct {
 	Type     string                   `json:"type"`
-	Response *OpenAIResponsesResponse `json:"response"`
+	Response *OpenAIResponsesResponse `json:"response,omitempty"`
 	Delta    string                   `json:"delta,omitempty"`
+	Item     *ResponsesOutput         `json:"item,omitempty"`
 }
--- a/relay/channel/openai/adaptor.go
+++ b/relay/channel/openai/adaptor.go
@@ -429,7 +429,7 @@ func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycom
 		if info.IsStream {
 			err, usage = OaiResponsesStreamHandler(c, resp, info)
 		} else {
-			err, usage = OpenaiResponsesHandler(c, resp, info)
+			err, usage = OaiResponsesHandler(c, resp, info)
 		}
 	default:
 		if info.IsStream {
--- a/relay/channel/openai/helper.go
+++ b/relay/channel/openai/helper.go
@@ -187,3 +187,10 @@ func handleFinalResponse(c *gin.Context, info *relaycommon.RelayInfo, lastStream
 		}
 	}
 }
+
+func sendResponsesStreamData(c *gin.Context, streamResponse dto.ResponsesStreamResponse, data string) {
+	if data == "" {
+		return
+	}
+	helper.ResponseChunkData(c, streamResponse, data)
+}
--- a/relay/channel/openai/relay-openai.go
+++ b/relay/channel/openai/relay-openai.go
@@ -644,102 +644,3 @@ func OpenaiHandlerWithUsage(c *gin.Context, resp *http.Response, info *relaycomm
 	}
 	return nil, &usageResp.Usage
 }
-
-func OpenaiResponsesHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
-	// read response body
-	var responsesResponse dto.OpenAIResponsesResponse
-	responseBody, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
-	}
-	err = resp.Body.Close()
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
-	}
-	err = common.DecodeJson(responseBody, &responsesResponse)
-	if err != nil {
-		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
-	}
-	if responsesResponse.Error != nil {
-		return &dto.OpenAIErrorWithStatusCode{
-			Error: dto.OpenAIError{
-				Message: responsesResponse.Error.Message,
-				Type:    "openai_error",
-				Code:    responsesResponse.Error.Code,
-			},
-			StatusCode: resp.StatusCode,
-		}, nil
-	}
-
-	// reset response body
-	resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
-	// We shouldn't set the header before we parse the response body, because the parse part may fail.
-	// And then we will have to send an error response, but in this case, the header has already been set.
-	// So the httpClient will be confused by the response.
-	// For example, Postman will report error, and we cannot check the response at all.
-	for k, v := range resp.Header {
-		c.Writer.Header().Set(k, v[0])
-	}
-	c.Writer.WriteHeader(resp.StatusCode)
-	// copy response body
-	_, err = io.Copy(c.Writer, resp.Body)
-	if err != nil {
-		common.SysError("error copying response body: " + err.Error())
-	}
-	resp.Body.Close()
-	// compute usage
-	usage := dto.Usage{}
-	usage.PromptTokens = responsesResponse.Usage.InputTokens
-	usage.CompletionTokens = responsesResponse.Usage.OutputTokens
-	usage.TotalTokens = responsesResponse.Usage.TotalTokens
-	return nil, &usage
-}
-
-func OaiResponsesStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
-	if resp == nil || resp.Body == nil {
-		common.LogError(c, "invalid response or response body")
-		return service.OpenAIErrorWrapper(fmt.Errorf("invalid response"), "invalid_response", http.StatusInternalServerError), nil
-	}
-
-	var usage = &dto.Usage{}
-	var responseTextBuilder strings.Builder
-
-	helper.StreamScannerHandler(c, resp, info, func(data string) bool {
-
-		// 检查当前数据是否包含 completed 状态和 usage 信息
-		var streamResponse dto.ResponsesStreamResponse
-		if err := common.DecodeJsonStr(data, &streamResponse); err == nil {
-			sendResponsesStreamData(c, streamResponse, data)
-			switch streamResponse.Type {
-			case "response.completed":
-				usage.PromptTokens = streamResponse.Response.Usage.InputTokens
-				usage.CompletionTokens = streamResponse.Response.Usage.OutputTokens
-				usage.TotalTokens = streamResponse.Response.Usage.TotalTokens
-			case "response.output_text.delta":
-				// 处理输出文本
-				responseTextBuilder.WriteString(streamResponse.Delta)
-
-			}
-		}
-		return true
-	})
-
-	if usage.CompletionTokens == 0 {
-		// 计算输出文本的 token 数量
-		tempStr := responseTextBuilder.String()
-		if len(tempStr) > 0 {
-			// 非正常结束，使用输出文本的 token 数量
-			completionTokens, _ := service.CountTextToken(tempStr, info.UpstreamModelName)
-			usage.CompletionTokens = completionTokens
-		}
-	}
-
-	return nil, usage
-}
-
-func sendResponsesStreamData(c *gin.Context, streamResponse dto.ResponsesStreamResponse, data string) {
-	if data == "" {
-		return
-	}
-	helper.ResponseChunkData(c, streamResponse, data)
-}
--- a/relay/channel/openai/relay_responses.go
+++ b/relay/channel/openai/relay_responses.go
@@ -0,0 +1,119 @@
+package openai
+
+import (
+	"bytes"
+	"fmt"
+	"io"
+	"net/http"
+	"one-api/common"
+	"one-api/dto"
+	relaycommon "one-api/relay/common"
+	"one-api/relay/helper"
+	"one-api/service"
+	"strings"
+
+	"github.com/gin-gonic/gin"
+)
+
+func OaiResponsesHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	// read response body
+	var responsesResponse dto.OpenAIResponsesResponse
+	responseBody, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = resp.Body.Close()
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil
+	}
+	err = common.DecodeJson(responseBody, &responsesResponse)
+	if err != nil {
+		return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil
+	}
+	if responsesResponse.Error != nil {
+		return &dto.OpenAIErrorWithStatusCode{
+			Error: dto.OpenAIError{
+				Message: responsesResponse.Error.Message,
+				Type:    "openai_error",
+				Code:    responsesResponse.Error.Code,
+			},
+			StatusCode: resp.StatusCode,
+		}, nil
+	}
+
+	// reset response body
+	resp.Body = io.NopCloser(bytes.NewBuffer(responseBody))
+	// We shouldn't set the header before we parse the response body, because the parse part may fail.
+	// And then we will have to send an error response, but in this case, the header has already been set.
+	// So the httpClient will be confused by the response.
+	// For example, Postman will report error, and we cannot check the response at all.
+	for k, v := range resp.Header {
+		c.Writer.Header().Set(k, v[0])
+	}
+	c.Writer.WriteHeader(resp.StatusCode)
+	// copy response body
+	_, err = io.Copy(c.Writer, resp.Body)
+	if err != nil {
+		common.SysError("error copying response body: " + err.Error())
+	}
+	resp.Body.Close()
+	// compute usage
+	usage := dto.Usage{}
+	usage.PromptTokens = responsesResponse.Usage.InputTokens
+	usage.CompletionTokens = responsesResponse.Usage.OutputTokens
+	usage.TotalTokens = responsesResponse.Usage.TotalTokens
+	// 解析 Tools 用量
+	for _, tool := range responsesResponse.Tools {
+		info.ResponsesUsageInfo.BuiltInTools[tool.Type].CallCount++
+	}
+	return nil, &usage
+}
+
+func OaiResponsesStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) {
+	if resp == nil || resp.Body == nil {
+		common.LogError(c, "invalid response or response body")
+		return service.OpenAIErrorWrapper(fmt.Errorf("invalid response"), "invalid_response", http.StatusInternalServerError), nil
+	}
+
+	var usage = &dto.Usage{}
+	var responseTextBuilder strings.Builder
+
+	helper.StreamScannerHandler(c, resp, info, func(data string) bool {
+
+		// 检查当前数据是否包含 completed 状态和 usage 信息
+		var streamResponse dto.ResponsesStreamResponse
+		if err := common.DecodeJsonStr(data, &streamResponse); err == nil {
+			sendResponsesStreamData(c, streamResponse, data)
+			switch streamResponse.Type {
+			case "response.completed":
+				usage.PromptTokens = streamResponse.Response.Usage.InputTokens
+				usage.CompletionTokens = streamResponse.Response.Usage.OutputTokens
+				usage.TotalTokens = streamResponse.Response.Usage.TotalTokens
+			case "response.output_text.delta":
+				// 处理输出文本
+				responseTextBuilder.WriteString(streamResponse.Delta)
+			case dto.ResponsesOutputTypeItemDone:
+				// 函数调用处理
+				if streamResponse.Item != nil {
+					switch streamResponse.Item.Type {
+					case dto.BuildInCallWebSearchCall:
+						info.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview].CallCount++
+					}
+				}
+			}
+		}
+		return true
+	})
+
+	if usage.CompletionTokens == 0 {
+		// 计算输出文本的 token 数量
+		tempStr := responseTextBuilder.String()
+		if len(tempStr) > 0 {
+			// 非正常结束，使用输出文本的 token 数量
+			completionTokens, _ := service.CountTextToken(tempStr, info.UpstreamModelName)
+			usage.CompletionTokens = completionTokens
+		}
+	}
+
+	return nil, usage
+}
--- a/relay/channel/vertex/adaptor.go
+++ b/relay/channel/vertex/adaptor.go
@@ -11,8 +11,8 @@ import (
 	"one-api/relay/channel/claude"
 	"one-api/relay/channel/gemini"
 	"one-api/relay/channel/openai"
-	"one-api/setting/model_setting"
 	relaycommon "one-api/relay/common"
+	"one-api/setting/model_setting"
 	"strings"

 	"github.com/gin-gonic/gin"
--- a/relay/common/relay_info.go
+++ b/relay/common/relay_info.go
@@ -36,6 +36,7 @@ type ClaudeConvertInfo struct {
 const (
 	RelayFormatOpenAI = "openai"
 	RelayFormatClaude = "claude"
+	RelayFormatGemini = "gemini"
 )

 type RerankerInfo struct {
@@ -43,6 +44,16 @@ type RerankerInfo struct {
 	ReturnDocuments bool
 }

+type BuildInToolInfo struct {
+	ToolName          string
+	CallCount         int
+	SearchContextSize string
+}
+
+type ResponsesUsageInfo struct {
+	BuiltInTools map[string]*BuildInToolInfo
+}
+
 type RelayInfo struct {
 	ChannelType       int
 	ChannelId         int
@@ -90,6 +101,7 @@ type RelayInfo struct {
 	ThinkingContentInfo
 	*ClaudeConvertInfo
 	*RerankerInfo
+	*ResponsesUsageInfo
 }

 // 定义支持流式选项的通道类型
@@ -134,6 +146,31 @@ func GenRelayInfoRerank(c *gin.Context, req *dto.RerankRequest) *RelayInfo {
 	return info
 }

+func GenRelayInfoResponses(c *gin.Context, req *dto.OpenAIResponsesRequest) *RelayInfo {
+	info := GenRelayInfo(c)
+	info.RelayMode = relayconstant.RelayModeResponses
+	info.ResponsesUsageInfo = &ResponsesUsageInfo{
+		BuiltInTools: make(map[string]*BuildInToolInfo),
+	}
+	if len(req.Tools) > 0 {
+		for _, tool := range req.Tools {
+			info.ResponsesUsageInfo.BuiltInTools[tool.Type] = &BuildInToolInfo{
+				ToolName:  tool.Type,
+				CallCount: 0,
+			}
+			switch tool.Type {
+			case dto.BuildInToolWebSearchPreview:
+				if tool.SearchContextSize == "" {
+					tool.SearchContextSize = "medium"
+				}
+				info.ResponsesUsageInfo.BuiltInTools[tool.Type].SearchContextSize = tool.SearchContextSize
+			}
+		}
+	}
+	info.IsStream = req.Stream
+	return info
+}
+
 func GenRelayInfo(c *gin.Context) *RelayInfo {
 	channelType := c.GetInt("channel_type")
 	channelId := c.GetInt("channel_id")
--- a/relay/relay-responses.go
+++ b/relay/relay-responses.go
@@ -19,7 +19,7 @@ import (
 	"github.com/gin-gonic/gin"
 )

-func getAndValidateResponsesRequest(c *gin.Context, relayInfo *relaycommon.RelayInfo) (*dto.OpenAIResponsesRequest, error) {
+func getAndValidateResponsesRequest(c *gin.Context) (*dto.OpenAIResponsesRequest, error) {
 	request := &dto.OpenAIResponsesRequest{}
 	err := common.UnmarshalBodyReusable(c, request)
 	if err != nil {
@@ -31,13 +31,11 @@ func getAndValidateResponsesRequest(c *gin.Context, relayInfo *relaycommon.Relay
 	if len(request.Input) == 0 {
 		return nil, errors.New("input is required")
 	}
-	relayInfo.IsStream = request.Stream
 	return request, nil

 }

 func checkInputSensitive(textRequest *dto.OpenAIResponsesRequest, info *relaycommon.RelayInfo) ([]string, error) {
-
 	sensitiveWords, err := service.CheckSensitiveInput(textRequest.Input)
 	return sensitiveWords, err
 }
@@ -49,12 +47,14 @@ func getInputTokens(req *dto.OpenAIResponsesRequest, info *relaycommon.RelayInfo
 }

 func ResponsesHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) {
-	relayInfo := relaycommon.GenRelayInfo(c)
-	req, err := getAndValidateResponsesRequest(c, relayInfo)
+	req, err := getAndValidateResponsesRequest(c)
 	if err != nil {
 		common.LogError(c, fmt.Sprintf("getAndValidateResponsesRequest error: %s", err.Error()))
 		return service.OpenAIErrorWrapperLocal(err, "invalid_responses_request", http.StatusBadRequest)
 	}
+
+	relayInfo := relaycommon.GenRelayInfoResponses(c, req)
+
 	if setting.ShouldCheckPromptSensitive() {
 		sensitiveWords, err := checkInputSensitive(req, relayInfo)
 		if err != nil {
--- a/relay/relay-text.go
+++ b/relay/relay-text.go
@@ -18,6 +18,7 @@ import (
 	"one-api/service"
 	"one-api/setting"
 	"one-api/setting/model_setting"
+	"one-api/setting/operation_setting"
 	"strings"
 	"time"

@@ -358,6 +359,34 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,

 	ratio := dModelRatio.Mul(dGroupRatio)

+	// openai web search 工具计费
+	var dWebSearchQuota decimal.Decimal
+	var webSearchPrice float64
+	if relayInfo.ResponsesUsageInfo != nil {
+		if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists && webSearchTool.CallCount > 0 {
+			// 计算 web search 调用的配额 (配额 = 价格 * 调用次数 / 1000)
+			webSearchPrice = operation_setting.GetWebSearchPricePerThousand(modelName, webSearchTool.SearchContextSize)
+			dWebSearchQuota = decimal.NewFromFloat(webSearchPrice).
+				Mul(decimal.NewFromInt(int64(webSearchTool.CallCount))).
+				Div(decimal.NewFromInt(1000))
+			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())
+		}
+	}
+	// file search tool 计费
+	var dFileSearchQuota decimal.Decimal
+	var fileSearchPrice float64
+	if relayInfo.ResponsesUsageInfo != nil {
+		if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists && fileSearchTool.CallCount > 0 {
+			fileSearchPrice = operation_setting.GetFileSearchPricePerThousand()
+			dFileSearchQuota = decimal.NewFromFloat(fileSearchPrice).
+				Mul(decimal.NewFromInt(int64(fileSearchTool.CallCount))).
+				Div(decimal.NewFromInt(1000))
+			extraContent += fmt.Sprintf("File Search 调用 %d 次，调用花费 $%s",
+				fileSearchTool.CallCount, dFileSearchQuota.String())
+		}
+	}
+
 	var quotaCalculateDecimal decimal.Decimal
 	if !priceData.UsePrice {
 		nonCachedTokens := dPromptTokens.Sub(dCacheTokens)
@@ -380,6 +409,9 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo,
 	} else {
 		quotaCalculateDecimal = dModelPrice.Mul(dQuotaPerUnit).Mul(dGroupRatio)
 	}
+	// 添加 responses tools call 调用的配额
+	quotaCalculateDecimal = quotaCalculateDecimal.Add(dWebSearchQuota)
+	quotaCalculateDecimal = quotaCalculateDecimal.Add(dFileSearchQuota)

 	quota := int(quotaCalculateDecimal.Round(0).IntPart())
 	totalTokens := promptTokens + completionTokens
@@ -430,6 +462,20 @@ 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 {
+			other["web_search"] = true
+			other["web_search_call_count"] = webSearchTool.CallCount
+			other["web_search_price"] = webSearchPrice
+		}
+	}
+	if !dFileSearchQuota.IsZero() && relayInfo.ResponsesUsageInfo != nil {
+		if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists {
+			other["file_search"] = true
+			other["file_search_call_count"] = fileSearchTool.CallCount
+			other["file_search_price"] = fileSearchPrice
+		}
+	}
 	model.RecordConsumeLog(ctx, relayInfo.UserId, relayInfo.ChannelId, promptTokens, completionTokens, logModel,
 		tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, relayInfo.Group, other)
 }
--- a/setting/operation_setting/tools.go
+++ b/setting/operation_setting/tools.go
@@ -0,0 +1,57 @@
+package operation_setting
+
+import "strings"
+
+const (
+	// Web search
+	WebSearchHighTierModelPriceLow    = 30.00
+	WebSearchHighTierModelPriceMedium = 35.00
+	WebSearchHighTierModelPriceHigh   = 50.00
+	WebSearchPriceLow                 = 25.00
+	WebSearchPriceMedium              = 27.50
+	WebSearchPriceHigh                = 30.00
+	// File search
+	FileSearchPrice = 2.5
+)
+
+func GetWebSearchPricePerThousand(modelName string, contextSize string) float64 {
+	// 确定模型类型
+	// https://platform.openai.com/docs/pricing Web search 价格按模型类型和 search context size 收费
+	// gpt-4.1, gpt-4o, or gpt-4o-search-preview 更贵，gpt-4.1-mini, gpt-4o-mini, gpt-4o-mini-search-preview 更便宜
+	isHighTierModel := (strings.HasPrefix(modelName, "gpt-4.1") || strings.HasPrefix(modelName, "gpt-4o")) &&
+		!strings.Contains(modelName, "mini")
+	// 确定 search context size 对应的价格
+	var priceWebSearchPerThousandCalls float64
+	switch contextSize {
+	case "low":
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceLow
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceLow
+		}
+	case "medium":
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceMedium
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceMedium
+		}
+	case "high":
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceHigh
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceHigh
+		}
+	default:
+		// search context size 默认为 medium
+		if isHighTierModel {
+			priceWebSearchPerThousandCalls = WebSearchHighTierModelPriceMedium
+		} else {
+			priceWebSearchPerThousandCalls = WebSearchPriceMedium
+		}
+	}
+	return priceWebSearchPerThousandCalls
+}
+
+func GetFileSearchPricePerThousand() float64 {
+	return FileSearchPrice
+}
--- a/web/src/components/LogsTable.js
+++ b/web/src/components/LogsTable.js
@@ -618,7 +618,6 @@ const LogsTable = () => {
            </Paragraph>
          );
        }
-
        let content = other?.claude
          ? renderClaudeModelPriceSimple(
              other.model_ratio,
@@ -935,6 +934,13 @@ const LogsTable = () => {
                other.model_price,
                other.group_ratio,
                other?.user_group_ratio,
+                false,
+                1.0,
+                undefined,
+                other.web_search || false,
+                other.web_search_call_count || 0,
+                other.file_search || false,
+                other.file_search_call_count || 0,
              ),
        });
      }
@@ -995,6 +1001,12 @@ const LogsTable = () => {
            other?.image || false,
            other?.image_ratio || 0,
            other?.image_output || 0,
+            other?.web_search || false,
+            other?.web_search_call_count || 0,
+            other?.web_search_price || 0,
+            other?.file_search || false,
+            other?.file_search_call_count || 0,
+            other?.file_search_price || 0,
          );
        }
        expandDataLocal.push({
--- a/web/src/helpers/render.js
+++ b/web/src/helpers/render.js
@@ -317,6 +317,12 @@ export function renderModelPrice(
  image = false,
  imageRatio = 1.0,
  imageOutputTokens = 0,
+  webSearch = false,
+  webSearchCallCount = 0,
+  webSearchPrice = 0,
+  fileSearch = false,
+  fileSearchCallCount = 0,
+  fileSearchPrice = 0,
 ) {
  if (modelPrice !== -1) {
    return i18next.t(
@@ -339,14 +345,17 @@ export function renderModelPrice(
    // Calculate effective input tokens (non-cached + cached with ratio applied)
    let effectiveInputTokens =
      inputTokens - cacheTokens + cacheTokens * cacheRatio;
-// Handle image tokens if present
+    // Handle image tokens if present
    if (image && imageOutputTokens > 0) {
-      effectiveInputTokens = inputTokens - imageOutputTokens + imageOutputTokens * imageRatio;
+      effectiveInputTokens =
+        inputTokens - imageOutputTokens + imageOutputTokens * imageRatio;
    }

    let price =
      (effectiveInputTokens / 1000000) * inputRatioPrice * groupRatio +
-      (completionTokens / 1000000) * completionRatioPrice * groupRatio;
+      (completionTokens / 1000000) * completionRatioPrice * groupRatio +
+      (webSearchCallCount / 1000) * webSearchPrice +
+      (fileSearchCallCount / 1000) * fileSearchPrice;

    return (
      <>
@@ -391,9 +400,23 @@ export function renderModelPrice(
              )}
            </p>
          )}
+          {webSearch && webSearchCallCount > 0 && (
+            <p>
+              {i18next.t('Web搜索价格：${{price}} / 1K 次', {
+                price: webSearchPrice,
+              })}
+            </p>
+          )}
+          {fileSearch && fileSearchCallCount > 0 && (
+            <p>
+              {i18next.t('文件搜索价格：${{price}} / 1K 次', {
+                price: fileSearchPrice,
+              })}
+            </p>
+          )}
          <p></p>
          <p>
-            {cacheTokens > 0 && !image
+            {cacheTokens > 0 && !image && !webSearch && !fileSearch
              ? i18next.t(
                  '输入 {{nonCacheInput}} tokens / 1M tokens * ${{price}} + 缓存 {{cacheInput}} tokens / 1M tokens * ${{cachePrice}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}',
                  {
@@ -407,31 +430,75 @@ export function renderModelPrice(
                    total: price.toFixed(6),
                  },
                )
-              : image && imageOutputTokens > 0
-              ? i18next.t(
-                  '输入 {{nonImageInput}} tokens + 图片输入 {{imageInput}} tokens * {{imageRatio}} / 1M tokens * ${{price}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}',
-                  {
-                    nonImageInput: inputTokens - imageOutputTokens,
-                    imageInput: imageOutputTokens,
-                    imageRatio: imageRatio,
-                    price: inputRatioPrice,
-                    completion: completionTokens,
-                    compPrice: completionRatioPrice,
-                    ratio: groupRatio,
-                    total: price.toFixed(6),
-                  },
-                )
-              : i18next.t(
-                  '输入 {{input}} tokens / 1M tokens * ${{price}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}',
-                  {
-                    input: inputTokens,
-                    price: inputRatioPrice,
-                    completion: completionTokens,
-                    compPrice: completionRatioPrice,
-                    ratio: groupRatio,
-                    total: price.toFixed(6),
-                  },
-                )}
+              : image && imageOutputTokens > 0 && !webSearch && !fileSearch
+                ? i18next.t(
+                    '输入 {{nonImageInput}} tokens + 图片输入 {{imageInput}} tokens * {{imageRatio}} / 1M tokens * ${{price}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}',
+                    {
+                      nonImageInput: inputTokens - imageOutputTokens,
+                      imageInput: imageOutputTokens,
+                      imageRatio: imageRatio,
+                      price: inputRatioPrice,
+                      completion: completionTokens,
+                      compPrice: completionRatioPrice,
+                      ratio: groupRatio,
+                      total: price.toFixed(6),
+                    },
+                  )
+                : webSearch && webSearchCallCount > 0 && !image && !fileSearch
+                  ? i18next.t(
+                      '输入 {{input}} tokens / 1M tokens * ${{price}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} + Web搜索 {{webSearchCallCount}}次 / 1K 次 * ${{webSearchPrice}} = ${{total}}',
+                      {
+                        input: inputTokens,
+                        price: inputRatioPrice,
+                        completion: completionTokens,
+                        compPrice: completionRatioPrice,
+                        ratio: groupRatio,
+                        webSearchCallCount,
+                        webSearchPrice,
+                        total: price.toFixed(6),
+                      },
+                    )
+                  : fileSearch && fileSearchCallCount > 0 && !image && !webSearch
+                    ? i18next.t(
+                        '输入 {{input}} tokens / 1M tokens * ${{price}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} + 文件搜索 {{fileSearchCallCount}}次 / 1K 次 * ${{fileSearchPrice}} = ${{total}}',
+                        {
+                          input: inputTokens,
+                          price: inputRatioPrice,
+                          completion: completionTokens,
+                          compPrice: completionRatioPrice,
+                          ratio: groupRatio,
+                          fileSearchCallCount,
+                          fileSearchPrice,
+                          total: price.toFixed(6),
+                        },
+                      )
+                    : webSearch && webSearchCallCount > 0 && fileSearch && fileSearchCallCount > 0 && !image
+                      ? i18next.t(
+                          '输入 {{input}} tokens / 1M tokens * ${{price}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} + Web搜索 {{webSearchCallCount}}次 / 1K 次 * ${{webSearchPrice}} + 文件搜索 {{fileSearchCallCount}}次 / 1K 次 * ${{fileSearchPrice}} = ${{total}}',
+                          {
+                            input: inputTokens,
+                            price: inputRatioPrice,
+                            completion: completionTokens,
+                            compPrice: completionRatioPrice,
+                            ratio: groupRatio,
+                            webSearchCallCount,
+                            webSearchPrice,
+                            fileSearchCallCount,
+                            fileSearchPrice,
+                            total: price.toFixed(6),
+                          },
+                        )
+                      : i18next.t(
+                          '输入 {{input}} tokens / 1M tokens * ${{price}} + 输出 {{completion}} tokens / 1M tokens * ${{compPrice}} * 分组 {{ratio}} = ${{total}}',
+                          {
+                            input: inputTokens,
+                            price: inputRatioPrice,
+                            completion: completionTokens,
+                            compPrice: completionRatioPrice,
+                            ratio: groupRatio,
+                            total: price.toFixed(6),
+                          },
+                        )}
          </p>
          <p>{i18next.t('仅供参考，以实际扣费为准')}</p>
        </article>
@@ -448,33 +515,56 @@ export function renderLogContent(
  user_group_ratio,
  image = false,
  imageRatio = 1.0,
-  useUserGroupRatio = undefined
+  useUserGroupRatio = undefined,
+  webSearch = false,
+  webSearchCallCount = 0,
+  fileSearch = false,
+  fileSearchCallCount = 0,
 ) {
-  const ratioLabel = useUserGroupRatio ? i18next.t('专属倍率') : i18next.t('分组倍率');
+  const ratioLabel = useUserGroupRatio
+    ? i18next.t('专属倍率')
+    : i18next.t('分组倍率');
  const ratio = useUserGroupRatio ? user_group_ratio : groupRatio;

  if (modelPrice !== -1) {
    return i18next.t('模型价格 ${{price}}，{{ratioType}} {{ratio}}', {
      price: modelPrice,
      ratioType: ratioLabel,
-      ratio
+      ratio,
    });
  } else {
    if (image) {
-      return i18next.t('模型倍率 {{modelRatio}}，输出倍率 {{completionRatio}}，图片输入倍率 {{imageRatio}}，{{ratioType}} {{ratio}}', {
-        modelRatio: modelRatio,
-        completionRatio: completionRatio,
-        imageRatio: imageRatio,
-        ratioType: ratioLabel,
-        ratio
-      });
+      return i18next.t(
+        '模型倍率 {{modelRatio}}，输出倍率 {{completionRatio}}，图片输入倍率 {{imageRatio}}，{{ratioType}} {{ratio}}',
+        {
+          modelRatio: modelRatio,
+          completionRatio: completionRatio,
+          imageRatio: imageRatio,
+          ratioType: ratioLabel,
+          ratio,
+        },
+      );
+    } else if (webSearch) {
+      return i18next.t(
+        '模型倍率 {{modelRatio}}，输出倍率 {{completionRatio}}，{{ratioType}} {{ratio}}，Web 搜索调用 {{webSearchCallCount}} 次',
+        {
+          modelRatio: modelRatio,
+          completionRatio: completionRatio,
+          ratioType: ratioLabel,
+          ratio,
+          webSearchCallCount,
+        },
+      );
    } else {
-      return i18next.t('模型倍率 {{modelRatio}}，输出倍率 {{completionRatio}}，{{ratioType}} {{ratio}}', {
-        modelRatio: modelRatio,
-        completionRatio: completionRatio,
-        ratioType: ratioLabel,
-        ratio
-      });
+      return i18next.t(
+        '模型倍率 {{modelRatio}}，输出倍率 {{completionRatio}}，{{ratioType}} {{ratio}}',
+        {
+          modelRatio: modelRatio,
+          completionRatio: completionRatio,
+          ratioType: ratioLabel,
+          ratio,
+        },
+      );
    }
  }
 }