From b0a7ce885e3659b1d75ce378dd4a08d77e21103a Mon Sep 17 00:00:00 2001
From: hhs <386998068@qq.com>
Date: Sat, 13 Jun 2026 21:09:56 +0800
Subject: [PATCH] =?UTF-8?q?fix:=20=E4=BF=AE=E5=A4=8D=E8=BF=94=E5=9B=9E?=
 =?UTF-8?q?=E7=BB=99=E5=89=8D=E7=AB=AF=E7=9A=84=20totalTokens=20=E4=B8=BA0?=
 =?UTF-8?q?=E7=9A=84=E9=97=AE=E9=A2=98=E5=B9=B6=E6=8F=90=E4=BE=9B=E7=A4=BA?=
 =?UTF-8?q?=E4=BE=8B=20.env?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

---
 .gitignore                                |  1 -
 backend/.env                              | 17 +++++++++++
 backend/.gitignore                        |  1 -
 backend/internal/orchestrator/pipeline.go | 36 ++++++++++++++++-------
 4 files changed, 43 insertions(+), 12 deletions(-)
 create mode 100644 backend/.env

diff --git a/.gitignore b/.gitignore
index 4c77b4c..d7a0342 100644
--- a/.gitignore
+++ b/.gitignore
@@ -7,7 +7,6 @@ backend/bin/
 backend/server
 
 # ---- 环境变量 ----
-.env
 .env.local
 .env.*.local
 
diff --git a/backend/.env b/backend/.env
new file mode 100644
index 0000000..48b24a0
--- /dev/null
+++ b/backend/.env
@@ -0,0 +1,17 @@
+# ---- AI 服务 API Key ----
+CAMTALK_AI_LLM_API_KEY=sk-ws-H.REHELLY.C4s3.MEUCIQCRee37XWEKp2szaxVLFDtR1rxNNsf372zMvCR0Xl6UvQIgZgvhRTvaa1FmhbCQJgaHu4Jny29AQkn01-3hX9CWBOg
+CAMTALK_AI_STT_API_KEY=tp-c9e7scwfx94qvqyhpnahnw8uaiya01za2qzvg4xe24rp3xiv
+CAMTALK_AI_TTS_API_KEY=tp-c9e7scwfx94qvqyhpnahnw8uaiya01za2qzvg4xe24rp3xiv
+
+# ---- 可选覆盖（默认值见 config.yaml）----
+# CAMTALK_AI_LLM_MODEL=qwen3-vl-plus
+# CAMTALK_AI_LLM_ENDPOINT=https://api.openai.com/v1
+# CAMTALK_AI_LLM_TIMEOUT=10
+# CAMTALK_AI_STT_ENDPOINT=wss://api.deepgram.com/v1/listen
+# CAMTALK_AI_TTS_ENDPOINT=https://api.openai.com/v1
+# CAMTALK_AI_TTS_VOICE=alloy
+# CAMTALK_AI_TTS_SPEED=1.0
+# CAMTALK_AI_TTS_TIMEOUT=5
+
+# ---- 应用 ----
+# APP_ENV=dev
diff --git a/backend/.gitignore b/backend/.gitignore
index 8304260..ac13b7b 100644
--- a/backend/.gitignore
+++ b/backend/.gitignore
@@ -3,7 +3,6 @@
 bin/
 
 # 环境配置
-.env
 config.dev.yaml
 config.prod.yaml
 
diff --git a/backend/internal/orchestrator/pipeline.go b/backend/internal/orchestrator/pipeline.go
index 0186108..d8d2823 100644
--- a/backend/internal/orchestrator/pipeline.go
+++ b/backend/internal/orchestrator/pipeline.go
@@ -166,11 +166,12 @@ func (p *Pipeline) ProcessQuery(
 	var ttsErr error
 
 	// goroutine 1: 消费 LLM token + 句子切分
+	var tokenUsage *llm.TokenUsage
 	wg.Add(1)
 	go func() {
 		defer wg.Done()
 		defer close(sentenceCh)
-		fullText = p.consumeLLMStream(ctx, llmStream, req.RequestID, sender, splitter)
+		fullText, tokenUsage = p.consumeLLMStream(ctx, llmStream, req.RequestID, sender, splitter)
 	}()
 
 	// goroutine 2: TTS 合成（如果启用）
@@ -205,13 +206,25 @@ func (p *Pipeline) ProcessQuery(
 
 	// 发送 llm_done
 	latency := time.Since(startTime).Milliseconds()
-	if err := sender.SendLLMDone(models.WsLLMDone{
+	done := models.WsLLMDone{
 		Type:      "llm_done",
 		RequestID: req.RequestID,
 		FullText:  fullText,
 		Model:     p.model,
 		LatencyMs: latency,
-	}); err != nil {
+	}
+	if tokenUsage != nil {
+		done.TokensUsed = struct {
+			Prompt     int `json:"prompt"`
+			Completion int `json:"completion"`
+			Total      int `json:"total"`
+		}{
+			Prompt:     tokenUsage.Prompt,
+			Completion: tokenUsage.Completion,
+			Total:      tokenUsage.Total,
+		}
+	}
+	if err := sender.SendLLMDone(done); err != nil {
 		log.Errorw("发送 llm_done 失败", "error", err)
 	}
 
@@ -225,33 +238,36 @@ func (p *Pipeline) ProcessQuery(
 }
 
 // consumeLLMStream 消费 LLM 流式输出，发送 llm_chunk 并进行句子切分。
+// 返回完整文本和 token 用量。
 func (p *Pipeline) consumeLLMStream(
 	ctx context.Context,
 	stream <-chan llm.Chunk,
 	requestID string,
 	sender Sender,
 	splitter *Splitter,
-) string {
+) (string, *llm.TokenUsage) {
 	log := logger.Log
 	var fullText strings.Builder
+	var tokenUsage *llm.TokenUsage
 
 	for chunk := range stream {
 		// 检查上下文是否已取消
 		select {
 		case <-ctx.Done():
 			log.Infow("LLM 流被中断", "request_id", requestID)
-			return fullText.String()
+			return fullText.String(), tokenUsage
 		default:
 		}
 
 		if chunk.Done {
-			// 流结束
+			// 流结束，记录 token 用量
 			if chunk.TokensUsed != nil {
+				tokenUsage = chunk.TokensUsed
 				log.Infow("LLM 用量统计",
 					"request_id", requestID,
-					"prompt_tokens", chunk.TokensUsed.Prompt,
-					"completion_tokens", chunk.TokensUsed.Completion,
-					"total_tokens", chunk.TokensUsed.Total,
+					"prompt_tokens", tokenUsage.Prompt,
+					"completion_tokens", tokenUsage.Completion,
+					"total_tokens", tokenUsage.Total,
 				)
 			}
 			break
@@ -277,7 +293,7 @@ func (p *Pipeline) consumeLLMStream(
 	// 刷新切分器中的剩余文本
 	splitter.Flush()
 
-	return fullText.String()
+	return fullText.String(), tokenUsage
 }
 
 // synthesizeTTS 从句子 channel 读取文本，进行 TTS 合成并发送音频。
-- 
2.49.1