add log

internal/controller/proxy/chat_proxy.go

@@ -60,22 +60,21 @@ func (h *Proxy) ChatHandler(c *gin.Context) {
 			c.SSEvent("", data)
 		}
 	}
-	go func() {
-		llmusage := llm.GetTokenUsage()
 
-		cost := tokenizer.Cost(llmusage.Model, llmusage.PromptTokens+llmusage.ToolsTokens, llmusage.CompletionTokens)
-		userid, _ := strconv.ParseInt(c.GetString("user_id"), 10, 64)
-		usage := model.Usage{
-			UserID:           userid,
-			Model:            llmusage.Model,
-			Stream:           chatreq.Stream,
-			PromptTokens:     llmusage.PromptTokens + llmusage.ToolsTokens,
-			CompletionTokens: llmusage.CompletionTokens,
-			TotalTokens:      llmusage.TotalTokens,
-			Cost:             fmt.Sprintf("%f", cost),
-		}
-		h.SendUsage(&usage)
-		defer fmt.Println("cost:", cost, "prompt_tokens:", llmusage.PromptTokens, "completion_tokens:", llmusage.CompletionTokens, "total_tokens:", llmusage.TotalTokens)
-	}()
+	llmusage := llm.GetTokenUsage()
+
+	cost := tokenizer.Cost(llmusage.Model, llmusage.PromptTokens+llmusage.ToolsTokens, llmusage.CompletionTokens)
+	userid, _ := strconv.ParseInt(c.GetString("user_id"), 10, 64)
+	usage := model.Usage{
+		UserID:           userid,
+		Model:            llmusage.Model,
+		Stream:           chatreq.Stream,
+		PromptTokens:     llmusage.PromptTokens + llmusage.ToolsTokens,
+		CompletionTokens: llmusage.CompletionTokens,
+		TotalTokens:      llmusage.TotalTokens,
+		Cost:             fmt.Sprintf("%f", cost),
+	}
+	h.SendUsage(&usage)
+	defer fmt.Println("cost:", cost, "prompt_tokens:", llmusage.PromptTokens, "completion_tokens:", llmusage.CompletionTokens, "total_tokens:", llmusage.TotalTokens)
 
 }

llm/llm.go

@@ -13,7 +13,7 @@ type LLM interface {
 
 type llm struct {
 	ApiKey   *model.ApiKey
-	Usage    *model.Usage
+	Usage    *TokenUsage
 	tools    any   // TODO
 	Messages []any // TODO
 	llm      LLM

llm/openai_compatible/chat.go

@@ -218,6 +218,7 @@ func (o *OpenAICompatible) StreamChat(ctx context.Context, chatReq llm.ChatReque
 			// case output <- &streamResp:
 			// }
 		}
+		fmt.Println("llm usage:", o.tokenUsage.Model, o.tokenUsage.PromptTokens, o.tokenUsage.CompletionTokens, o.tokenUsage.TotalTokens)
 	}()
 	return output, nil
 }