From 18b3300ff1d431a46a7cadf9a48db5c3c6ac0519 Mon Sep 17 00:00:00 2001 From: CaIon <1808837298@qq.com> Date: Mon, 5 May 2025 00:40:16 +0800 Subject: [PATCH 1/9] feat: implement OpenAI responses handling and streaming support with built-in tool tracking --- dto/openai_response.go | 11 ++- relay/channel/openai/adaptor.go | 2 +- relay/channel/openai/helper.go | 7 ++ relay/channel/openai/relay-openai.go | 99 -------------------- relay/channel/openai/relay_responses.go | 114 ++++++++++++++++++++++++ relay/channel/vertex/adaptor.go | 2 +- relay/common/relay_info.go | 37 ++++++++ relay/relay-responses.go | 10 +-- 8 files changed, 173 insertions(+), 109 deletions(-) create mode 100644 relay/channel/openai/relay_responses.go diff --git a/dto/openai_response.go b/dto/openai_response.go index 1508d1f6..c8f61b9d 100644 --- a/dto/openai_response.go +++ b/dto/openai_response.go @@ -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"` } diff --git a/relay/channel/openai/adaptor.go b/relay/channel/openai/adaptor.go index 7740c498..eb12a22a 100644 --- 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 { diff --git a/relay/channel/openai/helper.go b/relay/channel/openai/helper.go index e7ba2e7b..a068c544 100644 --- 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) +} diff --git a/relay/channel/openai/relay-openai.go b/relay/channel/openai/relay-openai.go index ef660564..b9ed94e2 100644 --- 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) -} diff --git a/relay/channel/openai/relay_responses.go b/relay/channel/openai/relay_responses.go new file mode 100644 index 00000000..6af8c676 --- /dev/null +++ b/relay/channel/openai/relay_responses.go @@ -0,0 +1,114 @@ +package openai + +import ( + "bytes" + "fmt" + "github.com/gin-gonic/gin" + "io" + "net/http" + "one-api/common" + "one-api/dto" + relaycommon "one-api/relay/common" + "one-api/relay/helper" + "one-api/service" + "strings" +) + +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 + 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 +} diff --git a/relay/channel/vertex/adaptor.go b/relay/channel/vertex/adaptor.go index c1b64f11..7daf9a61 100644 --- 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" diff --git a/relay/common/relay_info.go b/relay/common/relay_info.go index 915474e1..99c6d12b 100644 --- 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") diff --git a/relay/relay-responses.go b/relay/relay-responses.go index cdb37ae7..fd3ddb5a 100644 --- 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 { From b7fd1e4a203fb24d2b5a332ea4ec8abe3cdcecac Mon Sep 17 00:00:00 2001 From: tbphp Date: Mon, 5 May 2025 12:55:48 +0800 Subject: [PATCH 2/9] fix: Redis limit ignoring max eq 0 --- middleware/model-rate-limit.go | 36 ++++++++++++++++++---------------- 1 file changed, 19 insertions(+), 17 deletions(-) diff --git a/middleware/model-rate-limit.go b/middleware/model-rate-limit.go index 581dc451..f81160fc 100644 --- a/middleware/model-rate-limit.go +++ b/middleware/model-rate-limit.go @@ -93,25 +93,27 @@ func redisRateLimitHandler(duration int64, totalMaxCount, successMaxCount int) g } //2.检查总请求数限制并记录总请求（当totalMaxCount为0时会自动跳过，使用令牌桶限流器 - totalKey := fmt.Sprintf("rateLimit:%s", userId) - // 初始化 - tb := limiter.New(ctx, rdb) - allowed, err = tb.Allow( - ctx, - totalKey, - limiter.WithCapacity(int64(totalMaxCount)*duration), - limiter.WithRate(int64(totalMaxCount)), - limiter.WithRequested(duration), - ) + if totalMaxCount > 0 { + totalKey := fmt.Sprintf("rateLimit:%s", userId) + // 初始化 + tb := limiter.New(ctx, rdb) + allowed, err = tb.Allow( + ctx, + totalKey, + limiter.WithCapacity(int64(totalMaxCount)*duration), + limiter.WithRate(int64(totalMaxCount)), + limiter.WithRequested(duration), + ) - if err != nil { - fmt.Println("检查总请求数限制失败:", err.Error()) - abortWithOpenAiMessage(c, http.StatusInternalServerError, "rate_limit_check_failed") - return - } + if err != nil { + fmt.Println("检查总请求数限制失败:", err.Error()) + abortWithOpenAiMessage(c, http.StatusInternalServerError, "rate_limit_check_failed") + return + } - if !allowed { - abortWithOpenAiMessage(c, http.StatusTooManyRequests, fmt.Sprintf("您已达到总请求数限制：%d分钟内最多请求%d次，包括失败次数，请检查您的请求是否正确", setting.ModelRequestRateLimitDurationMinutes, totalMaxCount)) + if !allowed { + abortWithOpenAiMessage(c, http.StatusTooManyRequests, fmt.Sprintf("您已达到总请求数限制：%d分钟内最多请求%d次，包括失败次数，请检查您的请求是否正确", setting.ModelRequestRateLimitDurationMinutes, totalMaxCount)) + } } // 4. 处理请求 From 0cf4c59d227a90a8dd4b66927b7b563dc3cea72d Mon Sep 17 00:00:00 2001 From: skynono <6811626@qq.com> Date: Tue, 6 May 2025 14:18:15 +0800 Subject: [PATCH 3/9] feat: add original password verification when changing password --- controller/user.go | 26 +++++++++++++++++++++++++- model/user.go | 1 + web/src/components/PersonalSetting.js | 20 ++++++++++++++++++++ 3 files changed, 46 insertions(+), 1 deletion(-) diff --git a/controller/user.go b/controller/user.go index e194f531..567c2aa7 100644 --- a/controller/user.go +++ b/controller/user.go @@ -592,7 +592,14 @@ func UpdateSelf(c *gin.Context) { user.Password = "" // rollback to what it should be cleanUser.Password = "" } - updatePassword := user.Password != "" + updatePassword, err := checkUpdatePassword(user.OriginalPassword, user.Password, cleanUser.Id) + if err != nil { + c.JSON(http.StatusOK, gin.H{ + "success": false, + "message": err.Error(), + }) + return + } if err := cleanUser.Update(updatePassword); err != nil { c.JSON(http.StatusOK, gin.H{ "success": false, @@ -608,6 +615,23 @@ func UpdateSelf(c *gin.Context) { return } +func checkUpdatePassword(originalPassword string, newPassword string, userId int) (updatePassword bool, err error) { + if newPassword == "" { + return + } + var currentUser *model.User + currentUser, err = model.GetUserById(userId, true) + if err != nil { + return + } + if !common.ValidatePasswordAndHash(originalPassword, currentUser.Password) { + err = fmt.Errorf("原密码错误") + return + } + updatePassword = true + return +} + func DeleteUser(c *gin.Context) { id, err := strconv.Atoi(c.Param("id")) if err != nil { diff --git a/model/user.go b/model/user.go index 0aea2ff5..1a3372aa 100644 --- a/model/user.go +++ b/model/user.go @@ -18,6 +18,7 @@ type User struct { Id int `json:"id"` Username string `json:"username" gorm:"unique;index" validate:"max=12"` Password string `json:"password" gorm:"not null;" validate:"min=8,max=20"` + OriginalPassword string `json:"original_password" gorm:"-:all"` // this field is only for Password change verification, don't save it to database! DisplayName string `json:"display_name" gorm:"index" validate:"max=20"` Role int `json:"role" gorm:"type:int;default:1"` // admin, common Status int `json:"status" gorm:"type:int;default:1"` // enabled, disabled diff --git a/web/src/components/PersonalSetting.js b/web/src/components/PersonalSetting.js index d1e03db2..fbd74536 100644 --- a/web/src/components/PersonalSetting.js +++ b/web/src/components/PersonalSetting.js @@ -57,6 +57,7 @@ const PersonalSetting = () => { email_verification_code: '', email: '', self_account_deletion_confirmation: '', + original_password: '', set_new_password: '', set_new_password_confirmation: '', }); @@ -239,11 +240,20 @@ const PersonalSetting = () => { }; const changePassword = async () => { + if (inputs.original_password === '') { + showError(t('请输入原密码！')); + return; + } + if (inputs.original_password === inputs.set_new_password) { + showError(t('新密码需要和原密码不一致！')); + return; + } if (inputs.set_new_password !== inputs.set_new_password_confirmation) { showError(t('两次输入的密码不一致！')); return; } const res = await API.put(`/api/user/self`, { + original_password: inputs.original_password, password: inputs.set_new_password, }); const { success, message } = res.data; @@ -1118,6 +1128,16 @@ const PersonalSetting = () => { >
+ handleInputChange('original_password', value) + } + /> + Date: Tue, 6 May 2025 21:58:01 +0800 Subject: [PATCH 4/9] =?UTF-8?q?feat:=20=E6=B7=BB=E5=8A=A0=20built=20in=20t?= =?UTF-8?q?ools=20=E8=AE=A1=E8=B4=B9?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit - 增加非流的工具调用次数统计 - 添加 web search 和 file search 计费 --- dto/openai_response.go | 44 +++++++------- relay/channel/openai/relay_responses.go | 7 ++- relay/relay-text.go | 77 +++++++++++++++++++++++++ 3 files changed, 105 insertions(+), 23 deletions(-) diff --git a/dto/openai_response.go b/dto/openai_response.go index c8f61b9d..790d4df8 100644 --- 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 { diff --git a/relay/channel/openai/relay_responses.go b/relay/channel/openai/relay_responses.go index 6af8c676..1d1e060e 100644 --- a/relay/channel/openai/relay_responses.go +++ b/relay/channel/openai/relay_responses.go @@ -3,7 +3,6 @@ package openai import ( "bytes" "fmt" - "github.com/gin-gonic/gin" "io" "net/http" "one-api/common" @@ -12,6 +11,8 @@ import ( "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) { @@ -61,6 +62,10 @@ func OaiResponsesHandler(c *gin.Context, resp *http.Response, info *relaycommon. 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 } diff --git a/relay/relay-text.go b/relay/relay-text.go index 4fdd435d..a528ec52 100644 --- a/relay/relay-text.go +++ b/relay/relay-text.go @@ -358,6 +358,67 @@ func postConsumeQuota(ctx *gin.Context, relayInfo *relaycommon.RelayInfo, ratio := dModelRatio.Mul(dGroupRatio) + // openai web search 工具计费 + var dWebSearchQuota decimal.Decimal + if relayInfo.ResponsesUsageInfo != nil { + if webSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolWebSearchPreview]; exists && webSearchTool.CallCount > 0 { + // 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 webSearchTool.SearchContextSize { + case "low": + if isHighTierModel { + priceWebSearchPerThousandCalls = 30.0 + } else { + priceWebSearchPerThousandCalls = 25.0 + } + case "medium": + if isHighTierModel { + priceWebSearchPerThousandCalls = 35.0 + } else { + priceWebSearchPerThousandCalls = 27.5 + } + case "high": + if isHighTierModel { + priceWebSearchPerThousandCalls = 50.0 + } else { + priceWebSearchPerThousandCalls = 30.0 + } + default: + // search context size 默认为 medium + if isHighTierModel { + priceWebSearchPerThousandCalls = 35.0 + } else { + priceWebSearchPerThousandCalls = 27.5 + } + } + // 计算 web search 调用的配额 (配额 = 价格 * 调用次数 / 1000) + dWebSearchQuota = decimal.NewFromFloat(priceWebSearchPerThousandCalls). + 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 + if relayInfo.ResponsesUsageInfo != nil { + if fileSearchTool, exists := relayInfo.ResponsesUsageInfo.BuiltInTools[dto.BuildInToolFileSearch]; exists && fileSearchTool.CallCount > 0 { + // file search tool 调用价格 $2.50/1k calls + // 计算 file search tool 调用的配额 (配额 = 价格 * 调用次数 / 1000) + dFileSearchQuota = decimal.NewFromFloat(2.5). + 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 +441,9 @@ 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 +494,19 @@ 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_context_size"] = webSearchTool.SearchContextSize + } + } + 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 + } + } model.RecordConsumeLog(ctx, relayInfo.UserId, relayInfo.ChannelId, promptTokens, completionTokens, logModel, tokenName, quota, logContent, relayInfo.TokenId, userQuota, int(useTimeSeconds), relayInfo.IsStream, relayInfo.Group, other) } From d859e3fa645672ca3e38e97654a2de30c6bbd577 Mon Sep 17 00:00:00 2001 From: creamlike1024 Date: Tue, 6 May 2025 22:28:32 +0800 Subject: [PATCH 5/9] =?UTF-8?q?fix:=20=E4=BF=AE=E5=A4=8D=E6=9C=AA=E8=BE=93?= =?UTF-8?q?=E5=85=A5=E6=96=B0=E5=AF=86=E7=A0=81=E6=97=B6=E6=8F=90=E7=A4=BA?= =?UTF-8?q?=E4=BF=AE=E6=94=B9=E6=88=90=E5=8A=9F?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- controller/user.go | 6 ++--- web/src/components/PersonalSetting.js | 35 ++++++++++++++++++++------- 2 files changed, 29 insertions(+), 12 deletions(-) diff --git a/controller/user.go b/controller/user.go index 567c2aa7..fd53e743 100644 --- a/controller/user.go +++ b/controller/user.go @@ -616,9 +616,6 @@ func UpdateSelf(c *gin.Context) { } func checkUpdatePassword(originalPassword string, newPassword string, userId int) (updatePassword bool, err error) { - if newPassword == "" { - return - } var currentUser *model.User currentUser, err = model.GetUserById(userId, true) if err != nil { @@ -628,6 +625,9 @@ func checkUpdatePassword(originalPassword string, newPassword string, userId int err = fmt.Errorf("原密码错误") return } + if newPassword == "" { + return + } updatePassword = true return } diff --git a/web/src/components/PersonalSetting.js b/web/src/components/PersonalSetting.js index fbd74536..0f52c319 100644 --- a/web/src/components/PersonalSetting.js +++ b/web/src/components/PersonalSetting.js @@ -244,6 +244,10 @@ const PersonalSetting = () => { showError(t('请输入原密码！')); return; } + if (inputs.set_new_password === '') { + showError(t('请输入新密码！')); + return; + } if (inputs.original_password === inputs.set_new_password) { showError(t('新密码需要和原密码不一致！')); return; @@ -826,8 +830,8 @@ const PersonalSetting = () => {
- - + +
{t('通知方式')}
@@ -1003,23 +1007,36 @@ const PersonalSetting = () => {
- +
- {t('接受未设置价格模型')} + + {t('接受未设置价格模型')} +
handleNotificationSettingChange('acceptUnsetModelRatioModel', e.target.checked)} + checked={ + notificationSettings.acceptUnsetModelRatioModel + } + onChange={(e) => + handleNotificationSettingChange( + 'acceptUnsetModelRatioModel', + e.target.checked, + ) + } > {t('接受未设置价格模型')} - - {t('当模型没有设置价格时仍接受调用，仅当您信任该网站时使用，可能会产生高额费用')} + + {t( + '当模型没有设置价格时仍接受调用，仅当您信任该网站时使用，可能会产生高额费用', + )}
-