From d90e4bef63ac262bc3190002bab90180f69acdef Mon Sep 17 00:00:00 2001 From: creamlike1024 Date: Mon, 26 May 2025 14:50:50 +0800 Subject: [PATCH] gemini stream --- dto/gemini.go | 69 ------------------ relay/channel/gemini/adaptor.go | 7 +- relay/channel/gemini/relay-gemini-native.go | 81 +++++++++++++++++---- relay/relay-gemini.go | 28 +++++-- 4 files changed, 93 insertions(+), 92 deletions(-) delete mode 100644 dto/gemini.go diff --git a/dto/gemini.go b/dto/gemini.go deleted file mode 100644 index 898c966f..00000000 --- a/dto/gemini.go +++ /dev/null @@ -1,69 +0,0 @@ -package dto - -import "encoding/json" - -type GeminiPart struct { - Text string `json:"text"` -} - -type GeminiContent struct { - Parts []GeminiPart `json:"parts"` - Role string `json:"role"` -} - -type GeminiCandidate struct { - Content GeminiContent `json:"content"` - FinishReason string `json:"finishReason"` - AvgLogprobs float64 `json:"avgLogprobs"` -} - -type GeminiTokenDetails struct { - Modality string `json:"modality"` - TokenCount int `json:"tokenCount"` -} - -type GeminiUsageMetadata struct { - PromptTokenCount int `json:"promptTokenCount"` - CandidatesTokenCount int `json:"candidatesTokenCount"` - TotalTokenCount int `json:"totalTokenCount"` - PromptTokensDetails []GeminiTokenDetails `json:"promptTokensDetails"` - CandidatesTokensDetails []GeminiTokenDetails `json:"candidatesTokensDetails"` -} - -type GeminiTextGenerationResponse struct { - Candidates []GeminiCandidate `json:"candidates"` - UsageMetadata GeminiUsageMetadata `json:"usageMetadata"` - ModelVersion string `json:"modelVersion"` - ResponseID string `json:"responseId"` -} - -type GeminiGenerationConfig struct { - StopSequences []string `json:"stopSequences,omitempty"` - ResponseMimeType string `json:"responseMimeType,omitempty"` - ResponseSchema *json.RawMessage `json:"responseSchema,omitempty"` - ResponseModalities *json.RawMessage `json:"responseModalities,omitempty"` - CandidateCount int `json:"candidateCount,omitempty"` - MaxOutputTokens int `json:"maxOutputTokens,omitempty"` - Temperature float64 `json:"temperature,omitempty"` - TopP float64 `json:"topP,omitempty"` - TopK int `json:"topK,omitempty"` - Seed int `json:"seed,omitempty"` - PresencePenalty float64 `json:"presencePenalty,omitempty"` - FrequencyPenalty float64 `json:"frequencyPenalty,omitempty"` - ResponseLogprobs bool `json:"responseLogprobs,omitempty"` - LogProbs int `json:"logProbs,omitempty"` - EnableEnhancedCivicAnswers bool `json:"enableEnhancedCivicAnswers,omitempty"` - SpeechConfig *json.RawMessage `json:"speechConfig,omitempty"` - ThinkingConfig *json.RawMessage `json:"thinkingConfig,omitempty"` - MediaResolution *json.RawMessage `json:"mediaResolution,omitempty"` -} - -type GeminiTextGenerationRequest struct { - Contents []GeminiContent `json:"contents"` - Tools *json.RawMessage `json:"tools,omitempty"` - ToolConfig *json.RawMessage `json:"toolConfig,omitempty"` - SafetySettings *json.RawMessage `json:"safetySettings,omitempty"` - SystemInstruction *json.RawMessage `json:"systemInstruction,omitempty"` - GenerationConfig GeminiGenerationConfig `json:"generationConfig,omitempty"` - CachedContent *json.RawMessage `json:"cachedContent,omitempty"` -} diff --git a/relay/channel/gemini/adaptor.go b/relay/channel/gemini/adaptor.go index 12833736..e6f66d5f 100644 --- a/relay/channel/gemini/adaptor.go +++ b/relay/channel/gemini/adaptor.go @@ -167,8 +167,11 @@ func (a *Adaptor) DoRequest(c *gin.Context, info *relaycommon.RelayInfo, request func (a *Adaptor) DoResponse(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (usage any, err *dto.OpenAIErrorWithStatusCode) { if info.RelayMode == constant.RelayModeGemini { - err, usage = GeminiTextGenerationHandler(c, resp, info) - return usage, err + if info.IsStream { + return GeminiTextGenerationStreamHandler(c, resp, info) + } else { + return GeminiTextGenerationHandler(c, resp, info) + } } if strings.HasPrefix(info.UpstreamModelName, "imagen") { diff --git a/relay/channel/gemini/relay-gemini-native.go b/relay/channel/gemini/relay-gemini-native.go index 16374ea4..c055e299 100644 --- a/relay/channel/gemini/relay-gemini-native.go +++ b/relay/channel/gemini/relay-gemini-native.go @@ -7,20 +7,21 @@ import ( "one-api/common" "one-api/dto" relaycommon "one-api/relay/common" + "one-api/relay/helper" "one-api/service" "github.com/gin-gonic/gin" ) -func GeminiTextGenerationHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.OpenAIErrorWithStatusCode, *dto.Usage) { +func GeminiTextGenerationHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.Usage, *dto.OpenAIErrorWithStatusCode) { // 读取响应体 responseBody, err := io.ReadAll(resp.Body) if err != nil { - return service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError), nil + return nil, service.OpenAIErrorWrapper(err, "read_response_body_failed", http.StatusInternalServerError) } err = resp.Body.Close() if err != nil { - return service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError), nil + return nil, service.OpenAIErrorWrapper(err, "close_response_body_failed", http.StatusInternalServerError) } if common.DebugEnabled { @@ -28,15 +29,15 @@ func GeminiTextGenerationHandler(c *gin.Context, resp *http.Response, info *rela } // 解析为 Gemini 原生响应格式 - var geminiResponse dto.GeminiTextGenerationResponse + var geminiResponse GeminiChatResponse err = common.DecodeJson(responseBody, &geminiResponse) if err != nil { - return service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError), nil + return nil, service.OpenAIErrorWrapper(err, "unmarshal_response_body_failed", http.StatusInternalServerError) } // 检查是否有候选响应 if len(geminiResponse.Candidates) == 0 { - return &dto.OpenAIErrorWithStatusCode{ + return nil, &dto.OpenAIErrorWithStatusCode{ Error: dto.OpenAIError{ Message: "No candidates returned", Type: "server_error", @@ -44,7 +45,7 @@ func GeminiTextGenerationHandler(c *gin.Context, resp *http.Response, info *rela Code: 500, }, StatusCode: resp.StatusCode, - }, nil + } } // 计算使用量（基于 UsageMetadata） @@ -54,15 +55,10 @@ func GeminiTextGenerationHandler(c *gin.Context, resp *http.Response, info *rela TotalTokens: geminiResponse.UsageMetadata.TotalTokenCount, } - // 设置模型版本 - if geminiResponse.ModelVersion == "" { - geminiResponse.ModelVersion = info.UpstreamModelName - } - // 直接返回 Gemini 原生格式的 JSON 响应 jsonResponse, err := json.Marshal(geminiResponse) if err != nil { - return service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError), nil + return nil, service.OpenAIErrorWrapper(err, "marshal_response_body_failed", http.StatusInternalServerError) } // 设置响应头并写入响应 @@ -70,8 +66,63 @@ func GeminiTextGenerationHandler(c *gin.Context, resp *http.Response, info *rela c.Writer.WriteHeader(resp.StatusCode) _, err = c.Writer.Write(jsonResponse) if err != nil { - return service.OpenAIErrorWrapper(err, "write_response_failed", http.StatusInternalServerError), nil + return nil, service.OpenAIErrorWrapper(err, "write_response_failed", http.StatusInternalServerError) } - return nil, &usage + return &usage, nil +} + +func GeminiTextGenerationStreamHandler(c *gin.Context, resp *http.Response, info *relaycommon.RelayInfo) (*dto.Usage, *dto.OpenAIErrorWithStatusCode) { + var usage = &dto.Usage{} + var imageCount int + + helper.SetEventStreamHeaders(c) + + helper.StreamScannerHandler(c, resp, info, func(data string) bool { + var geminiResponse GeminiChatResponse + err := common.DecodeJsonStr(data, &geminiResponse) + if err != nil { + common.LogError(c, "error unmarshalling stream response: "+err.Error()) + return false + } + + // 统计图片数量 + for _, candidate := range geminiResponse.Candidates { + for _, part := range candidate.Content.Parts { + if part.InlineData != nil && part.InlineData.MimeType != "" { + imageCount++ + } + } + } + + // 更新使用量统计 + if geminiResponse.UsageMetadata.TotalTokenCount != 0 { + usage.PromptTokens = geminiResponse.UsageMetadata.PromptTokenCount + usage.CompletionTokens = geminiResponse.UsageMetadata.CandidatesTokenCount + usage.TotalTokens = geminiResponse.UsageMetadata.TotalTokenCount + } + + // 直接发送 GeminiChatResponse 响应 + err = helper.ObjectData(c, geminiResponse) + if err != nil { + common.LogError(c, err.Error()) + } + + return true + }) + + if imageCount != 0 { + if usage.CompletionTokens == 0 { + usage.CompletionTokens = imageCount * 258 + } + } + + // 计算最终使用量 + usage.PromptTokensDetails.TextTokens = usage.PromptTokens + usage.CompletionTokens = usage.TotalTokens - usage.PromptTokens + + // 结束流式响应 + helper.Done(c) + + return usage, nil } diff --git a/relay/relay-gemini.go b/relay/relay-gemini.go index 9aa072e1..93a2b7aa 100644 --- a/relay/relay-gemini.go +++ b/relay/relay-gemini.go @@ -8,6 +8,7 @@ import ( "net/http" "one-api/common" "one-api/dto" + "one-api/relay/channel/gemini" relaycommon "one-api/relay/common" "one-api/relay/helper" "one-api/service" @@ -17,8 +18,8 @@ import ( "github.com/gin-gonic/gin" ) -func getAndValidateGeminiRequest(c *gin.Context) (*dto.GeminiTextGenerationRequest, error) { - request := &dto.GeminiTextGenerationRequest{} +func getAndValidateGeminiRequest(c *gin.Context) (*gemini.GeminiChatRequest, error) { + request := &gemini.GeminiChatRequest{} err := common.UnmarshalBodyReusable(c, request) if err != nil { return nil, err @@ -29,7 +30,19 @@ func getAndValidateGeminiRequest(c *gin.Context) (*dto.GeminiTextGenerationReque return request, nil } -func checkGeminiInputSensitive(textRequest *dto.GeminiTextGenerationRequest, info *relaycommon.RelayInfo) ([]string, error) { +// 流模式 +// /v1beta/models/gemini-2.0-flash:streamGenerateContent?alt=sse&key=xxx +func checkGeminiStreamMode(c *gin.Context, relayInfo *relaycommon.RelayInfo) { + if c.Query("alt") == "sse" { + relayInfo.IsStream = true + } + + // if strings.Contains(c.Request.URL.Path, "streamGenerateContent") { + // relayInfo.IsStream = true + // } +} + +func checkGeminiInputSensitive(textRequest *gemini.GeminiChatRequest) ([]string, error) { var inputTexts []string for _, content := range textRequest.Contents { for _, part := range content.Parts { @@ -46,7 +59,7 @@ func checkGeminiInputSensitive(textRequest *dto.GeminiTextGenerationRequest, inf return sensitiveWords, err } -func getGeminiInputTokens(req *dto.GeminiTextGenerationRequest, info *relaycommon.RelayInfo) (int, error) { +func getGeminiInputTokens(req *gemini.GeminiChatRequest, info *relaycommon.RelayInfo) (int, error) { // 计算输入 token 数量 var inputTexts []string for _, content := range req.Contents { @@ -72,8 +85,11 @@ func GeminiHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) { relayInfo := relaycommon.GenRelayInfo(c) + // 检查 Gemini 流式模式 + checkGeminiStreamMode(c, relayInfo) + if setting.ShouldCheckPromptSensitive() { - sensitiveWords, err := checkGeminiInputSensitive(req, relayInfo) + sensitiveWords, err := checkGeminiInputSensitive(req) if err != nil { common.LogWarn(c, fmt.Sprintf("user sensitive words detected: %s", strings.Join(sensitiveWords, ", "))) return service.OpenAIErrorWrapperLocal(err, "check_request_sensitive_error", http.StatusBadRequest) @@ -97,7 +113,7 @@ func GeminiHelper(c *gin.Context) (openaiErr *dto.OpenAIErrorWithStatusCode) { c.Set("prompt_tokens", promptTokens) } - priceData, err := helper.ModelPriceHelper(c, relayInfo, relayInfo.PromptTokens, req.GenerationConfig.MaxOutputTokens) + priceData, err := helper.ModelPriceHelper(c, relayInfo, relayInfo.PromptTokens, int(req.GenerationConfig.MaxOutputTokens)) if err != nil { return service.OpenAIErrorWrapperLocal(err, "model_price_error", http.StatusInternalServerError) }