diff --git a/monitor/manage.go b/monitor/manage.go index 44c13612..268d3924 100644 --- a/monitor/manage.go +++ b/monitor/manage.go @@ -34,7 +34,7 @@ func ShouldDisableChannel(err *model.Error, statusCode int) bool { strings.Contains(lowerMessage, "credit") || strings.Contains(lowerMessage, "balance") || strings.Contains(lowerMessage, "permission denied") || - strings.Contains(lowerMessage, "organization has been restricted") || // groq + strings.Contains(lowerMessage, "organization has been restricted") || // groq strings.Contains(lowerMessage, "已欠费") { return true } diff --git a/relay/adaptor/ollama/main.go b/relay/adaptor/ollama/main.go index 39549f2a..2462fc3e 100644 --- a/relay/adaptor/ollama/main.go +++ b/relay/adaptor/ollama/main.go @@ -30,8 +30,8 @@ func ConvertRequest(request model.GeneralOpenAIRequest) *ChatRequest { TopP: request.TopP, FrequencyPenalty: request.FrequencyPenalty, PresencePenalty: request.PresencePenalty, - NumPredict: request.MaxTokens, - NumCtx: request.NumCtx, + NumPredict: request.MaxTokens, + NumCtx: request.NumCtx, }, Stream: request.Stream, } @@ -121,7 +121,7 @@ func StreamHandler(c *gin.Context, resp *http.Response) (*model.ErrorWithStatusC for scanner.Scan() { data := scanner.Text() if strings.HasPrefix(data, "}") { - data = strings.TrimPrefix(data, "}") + "}" + data = strings.TrimPrefix(data, "}") + "}" } var ollamaResponse ChatResponse diff --git a/relay/billing/ratio/model.go b/relay/billing/ratio/model.go index 1b58ec09..b8f140a1 100644 --- a/relay/billing/ratio/model.go +++ b/relay/billing/ratio/model.go @@ -83,6 +83,7 @@ var ModelRatio = map[string]float64{ "claude-3-sonnet-20240229": 3.0 / 1000 * USD, "claude-3-5-sonnet-20240620": 3.0 / 1000 * USD, "claude-3-5-sonnet-20241022": 3.0 / 1000 * USD, + "claude-3-5-sonnet-latest": 3.0 / 1000 * USD, "claude-3-opus-20240229": 15.0 / 1000 * USD, // https://cloud.baidu.com/doc/WENXINWORKSHOP/s/hlrk4akp7 "ERNIE-4.0-8K": 0.120 * RMB,