@xdfnet/ispeak 1.6.6 → 1.6.7

package/Docs/HOOK_TEXT_EXTRACTION.md

@@ -205,6 +205,7 @@ fi
 
 - Codex `notify`：读 `$2`
 - Claude / Codex Stop Hook：读 stdin
+- 如果 Codex 的 `notify` 和 `Stop` 同时启用，脚本会按 `turn_id` 去重，避免同一回合播两次
 
 Codex 文本字段优先级：
 

package/README.md

@@ -1,6 +1,6 @@
 # iSpeak
 
-![Version](https://img.shields.io/badge/version-1.6.6-blue)
+![Version](https://img.shields.io/badge/version-1.6.7-blue)
 [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT)
 [![Go Version](https://img.shields.io/badge/Go-1.26-blue)](https://golang.org/dl/)
 ![Platform](https://img.shields.io/badge/platform-macOS-green)
@@ -39,11 +39,7 @@ ispeak-codex "构建完成，耗时 12 秒"
 npm i -g @xdfnet/ispeak
 ```
 
-当前 npm 安装会在本机编译 `ispeakd`，需要已安装 Go。没有 `ffplay` 时会自动回退 `afplay`；推荐安装 `ffmpeg` 获得流式播放：
-
-```bash
-brew install ffmpeg
-```
+当前 npm 安装会在本机编译 `ispeakd`，需要已安装 Go。主播放链路使用 macOS 原生 `AVAudioEngine`，不依赖 `ffmpeg`。失败时直接记录日志并删除任务。
 
 **源码安装：**
 
@@ -76,11 +72,11 @@ ispeak "iSpeak 准备好了"
 │         │                                            │
 │         ▼                                           │
 │   单 Worker 流式链路                                 │
-│   （SSE audio chunk → 播放器 stdin）                  │
+│   （SSE PCM chunk → AVAudioEngine）                 │
 │         │                                            │
 │         ▼                                           │
-│   流式播放器                                         │
-│   （优先 ffplay stdin，无 ffplay 回退 afplay）         │
+│   错误处理                                          │
+│   （失败时记录日志并删除任务）                        │
 └─────────────────────────────────────────────────────┘
 ```
 
@@ -194,6 +190,8 @@ notify = ["bash", "/Users/你的用户名/.config/iSpeak/hook-speak.sh", "codex"
 }
 ```
 
+`hook-speak.sh` 会按 `turn_id` 做一次去重，所以即使 `notify` 和 `Stop` 都启用，同一回合也只会播一次。
+
 ## 开发命令
 
 ```bash

package/configs/hook-speak.sh

@@ -6,6 +6,7 @@
 SOURCE="${1:-claude}"
 SOCK="$HOME/.config/iSpeak/ispeak.sock"
 LOG="$HOME/.config/iSpeak/hook.log"
+STATE_FILE="$HOME/.config/iSpeak/hook.last"
 
 # Codex `notify` 会把 JSON 作为最后一个参数传入；
 # Claude/Claude 风格 Stop Hook 会把 JSON 写到 stdin。
@@ -17,26 +18,41 @@ input_file=$(mktemp)
 trap 'rm -f "$input_file"' EXIT
 printf "%s" "$input" > "$input_file"
 
-text=$(SOURCE="$SOURCE" HOOK_INPUT_FILE="$input_file" node <<'NODE' 2>/dev/null
+result=$(SOURCE="$SOURCE" HOOK_INPUT_FILE="$input_file" HOOK_STATE_FILE="$STATE_FILE" node <<'NODE' 2>/dev/null
 const fs = require("fs");
+const crypto = require("crypto");
 
-{
+(() => {
   const input = readFile(process.env.HOOK_INPUT_FILE || "");
   const payload = parseJSON(input) || {};
   const source = process.env.SOURCE || "";
-  const text = source.startsWith("codex")
+  const stateFile = process.env.HOOK_STATE_FILE || "";
+  const result = source.startsWith("codex")
     ? lastCodexAssistant(payload)
     : lastClaudeAssistant(payload);
 
-  if (text) process.stdout.write(text);
-}
+  if (!result.text) {
+    return;
+  }
+
+  if (stateFile && result.turnId) {
+    if (isDuplicateTurn(stateFile, source, result.turnId)) {
+      return;
+    }
+    saveTurn(stateFile, source, result.turnId, result.text);
+  } else if (stateFile) {
+    saveTurn(stateFile, source, "", result.text);
+  }
+
+  process.stdout.write(result.text);
+})();
 
 function lastClaudeAssistant(payload) {
   const direct = firstString(payload.last_assistant_message, payload.message);
-  if (direct) return direct;
+  if (direct) return { text: direct, turnId: extractTurnId(payload) };
 
   const transcript = firstString(payload.transcript_path, payload.transcriptPath);
-  return transcript ? lastAssistantFromTranscript(transcript, "claude") : "";
+  return transcript ? lastAssistantFromTranscript(transcript, "claude") : { text: "", turnId: extractTurnId(payload) };
 }
 
 function lastCodexAssistant(payload) {
@@ -47,14 +63,14 @@ function lastCodexAssistant(payload) {
     payload.message,
     payload.lastMessage
   );
-  if (direct) return direct;
+  if (direct) return { text: direct, turnId: extractTurnId(payload) };
 
   const transcript = firstString(
     payload.transcript_path,
     payload.transcriptPath,
     payload["transcript-path"]
   );
-  return transcript ? lastAssistantFromTranscript(transcript, "codex") : "";
+  return transcript ? lastAssistantFromTranscript(transcript, "codex") : { text: "", turnId: extractTurnId(payload) };
 }
 
 function readFile(file) {
@@ -98,6 +114,7 @@ function lastAssistantFromTranscript(file, source) {
   }
 
   let last = "";
+  let turnId = "";
   for (const line of data.split(/\r?\n/)) {
     if (!line.trim()) continue;
     const event = parseJSON(line);
@@ -119,26 +136,64 @@ function lastAssistantFromTranscript(file, source) {
       event.payload.role === "assistant"
     ) {
       last = collectText(event.payload.content) || last;
+      turnId = turnId || extractTurnId(event) || extractTurnId(event.payload);
     }
   }
-  return last;
+  return { text: last, turnId };
+}
+
+function extractTurnId(payload) {
+  return firstString(
+    payload.turn_id,
+    payload.turnId,
+    payload["turn-id"],
+    payload.session_id,
+    payload.sessionId,
+    payload["session-id"],
+    payload.thread_id,
+    payload.threadId,
+    payload["thread-id"]
+  );
+}
+
+function isDuplicateTurn(stateFile, source, turnId) {
+  const current = `${source}:${turnId}`;
+  try {
+    return fs.readFileSync(stateFile, "utf8").trim() === current;
+  } catch {
+    return false;
+  }
+}
+
+function saveTurn(stateFile, source, turnId, text) {
+  const current = `${source}:${turnId || textHash(text)}`;
+  try {
+    fs.mkdirSync(require("path").dirname(stateFile), { recursive: true });
+    fs.writeFileSync(stateFile, current, "utf8");
+  } catch {
+    // 去重失败不影响播报。
+  }
+}
+
+function textHash(text) {
+  return crypto.createHash("sha1").update(text, "utf8").digest("hex");
 }
 
 NODE
 )
 
 if [[ "$ISPEAK_HOOK_PRINT_TEXT" == "1" ]]; then
-  printf "%s" "$text"
+  printf "%s" "$result"
   exit 0
 fi
 
 echo "=== $(date) ===" >> "$LOG"
 echo "SOURCE: $SOURCE" >> "$LOG"
-echo "TEXT_LEN: ${#text}" >> "$LOG"
-echo "PREVIEW: ${text:0:150}" >> "$LOG"
+echo "TEXT_LEN: ${#result}" >> "$LOG"
+echo "PREVIEW: ${result:0:150}" >> "$LOG"
 
-if [[ -n "$text" && -S "$SOCK" ]]; then
-  printf "{source:%s}%s" "$SOURCE" "$text" | nc -U -w5 "$SOCK" 2>> "$LOG"
+if [[ -n "$result" && -S "$SOCK" ]]; then
+  printf "{source:%s}%s" "$SOURCE" "$result" | nc -U -w5 "$SOCK" 2>> "$LOG"
   echo "SPOKE: OK" >> "$LOG"
 else
   echo "SPOKE: SKIP" >> "$LOG"

package/main.go

@@ -1,5 +1,5 @@
 // ttsd — 独立 TTS 播报守护进程
-// 监听 Unix Socket，收到文本 → 字节跳动 TTS SSE → 流式播放
+// 监听 Unix Socket，收到文本 → 字节跳动 TTS SSE/PCM → 原生流式播放
 package main
 
 import (
@@ -14,7 +14,6 @@ import (
 	"net"
 	"net/http"
 	"os"
-	"os/exec"
 	"os/signal"
 	"path/filepath"
 	"strings"
@@ -48,137 +47,6 @@ type StreamPlayer interface {
 	Abort() error
 }
 
-type ffplayStreamPlayer struct {
-	path string
-	cmd  *exec.Cmd
-
-	mu       sync.Mutex
-	stdin    io.WriteCloser
-	waitOnce sync.Once
-	waitErr  error
-}
-
-func newDefaultStreamPlayer() (StreamPlayer, error) {
-	if path, ok := findExecutable("ffplay", "/opt/homebrew/bin/ffplay", "/usr/local/bin/ffplay"); ok {
-		log.Printf("播放器模式: ffplay 流式 stdin (%s)", path)
-		return newFFplayStreamPlayer(path)
-	}
-
-	log.Printf("播放器模式: afplay 完整音频 fallback")
-	return &bufferedStreamPlayer{}, nil
-}
-
-func findExecutable(name string, candidates ...string) (string, bool) {
-	if path, err := exec.LookPath(name); err == nil {
-		return path, true
-	}
-	for _, path := range candidates {
-		if st, err := os.Stat(path); err == nil && !st.IsDir() && st.Mode()&0111 != 0 {
-			return path, true
-		}
-	}
-	return "", false
-}
-
-func newFFplayStreamPlayer(path string) (*ffplayStreamPlayer, error) {
-	cmd := exec.Command(path, "-nodisp", "-autoexit", "-loglevel", "error", "-i", "pipe:0")
-	stdin, err := cmd.StdinPipe()
-	if err != nil {
-		return nil, err
-	}
-	cmd.Stderr = os.Stderr
-	if err := cmd.Start(); err != nil {
-		_ = stdin.Close()
-		return nil, err
-	}
-	return &ffplayStreamPlayer{path: path, cmd: cmd, stdin: stdin}, nil
-}
-
-func (p *ffplayStreamPlayer) Write(audio []byte) error {
-	if len(audio) == 0 {
-		return nil
-	}
-	p.mu.Lock()
-	stdin := p.stdin
-	p.mu.Unlock()
-	if stdin == nil {
-		return fmt.Errorf("播放器输入已关闭")
-	}
-	if _, err := stdin.Write(audio); err != nil {
-		return fmt.Errorf("写入播放器失败: %w", err)
-	}
-	return nil
-}
-
-func (p *ffplayStreamPlayer) CloseAndWait() error {
-	p.mu.Lock()
-	stdin := p.stdin
-	p.stdin = nil
-	p.mu.Unlock()
-	if stdin != nil {
-		if err := stdin.Close(); err != nil {
-			return fmt.Errorf("关闭播放器输入失败: %w", err)
-		}
-	}
-	if err := p.wait(); err != nil {
-		return fmt.Errorf("ffplay failed: %w", err)
-	}
-	return nil
-}
-
-func (p *ffplayStreamPlayer) Abort() error {
-	p.mu.Lock()
-	stdin := p.stdin
-	p.stdin = nil
-	p.mu.Unlock()
-	if stdin != nil {
-		_ = stdin.Close()
-	}
-	if p.cmd != nil && p.cmd.Process != nil {
-		_ = p.cmd.Process.Kill()
-	}
-	return p.wait()
-}
-
-func (p *ffplayStreamPlayer) wait() error {
-	p.waitOnce.Do(func() {
-		if p.cmd != nil {
-			p.waitErr = p.cmd.Wait()
-		}
-	})
-	return p.waitErr
-}
-
-type bufferedStreamPlayer struct {
-	chunks [][]byte
-}
-
-func (p *bufferedStreamPlayer) Write(audio []byte) error {
-	if len(audio) == 0 {
-		return nil
-	}
-	chunk := append([]byte(nil), audio...)
-	p.chunks = append(p.chunks, chunk)
-	return nil
-}
-
-func (p *bufferedStreamPlayer) CloseAndWait() error {
-	total := 0
-	for _, chunk := range p.chunks {
-		total += len(chunk)
-	}
-	audio := make([]byte, 0, total)
-	for _, chunk := range p.chunks {
-		audio = append(audio, chunk...)
-	}
-	return playAudio(audio)
-}
-
-func (p *bufferedStreamPlayer) Abort() error {
-	p.chunks = nil
-	return nil
-}
-
 // 任务状态
 // 生命周期：pending -> running -> delete
 type TaskStatus int
@@ -312,13 +180,13 @@ func (e *TaskEngine) runTransaction(task *Task) error {
 
 	if err := e.synthesizeStreamFn(context.Background(), task.Cfg, task.Text, &task.Voice, onAudio); err != nil {
 		_ = player.Abort()
-		return err
+		return fmt.Errorf("TTS 合成失败: id=%d: %w", task.ID, err)
 	}
 	log.Printf("TTS 流结束: id=%d elapsed=%s", task.ID, time.Since(startedAt).Round(time.Millisecond))
 
 	if err := player.CloseAndWait(); err != nil {
 		_ = player.Abort()
-		return err
+		return fmt.Errorf("播放器失败: id=%d: %w", task.ID, err)
 	}
 	return nil
 }
@@ -476,28 +344,6 @@ type ttsAudioParams struct {
 	SampleRate int    `json:"sample_rate"`
 }
 
-// 调用字节跳动 TTS API，返回完整 MP3 音频数据。保留给测试和 fallback 使用。
-func synthesize(ctx context.Context, cfg Config, text string, voice *VoiceInfo) ([]byte, error) {
-	var chunks [][]byte
-	if err := synthesizeStream(ctx, cfg, text, voice, func(audio []byte) error {
-		chunk := append([]byte(nil), audio...)
-		chunks = append(chunks, chunk)
-		return nil
-	}); err != nil {
-		return nil, err
-	}
-
-	total := 0
-	for _, c := range chunks {
-		total += len(c)
-	}
-	result := make([]byte, 0, total)
-	for _, c := range chunks {
-		result = append(result, c...)
-	}
-	return result, nil
-}
-
 // 调用字节跳动 TTS API，边解析 SSE 边回调 MP3 音频块
 func synthesizeStream(ctx context.Context, cfg Config, text string, voice *VoiceInfo, onAudio func([]byte) error) error {
 	speaker := voice.VoiceType
@@ -512,8 +358,8 @@ func synthesizeStream(ctx context.Context, cfg Config, text string, voice *Voice
 			Text:    text,
 			Speaker: speaker,
 			AudioParams: ttsAudioParams{
-				Format:     "mp3",
-				SampleRate: 24000,
+				Format:     "pcm",
+				SampleRate: 48000,
 			},
 		},
 	}
@@ -848,21 +694,6 @@ func validateVoiceInfo(name string, voice *VoiceInfo) error {
 	return nil
 }
 
-func playAudio(data []byte) error {
-	tmpFile := filepath.Join(tempDir, fmt.Sprintf("ttsd-%d.mp3", time.Now().UnixNano()))
-	if err := os.WriteFile(tmpFile, data, 0644); err != nil {
-		return fmt.Errorf("写入临时文件失败: %w", err)
-	}
-	defer os.Remove(tmpFile)
-
-	cmd := exec.Command("/usr/bin/afplay", tmpFile)
-	log.Printf("播放开始: %s", filepath.Base(tmpFile))
-	if err := cmd.Run(); err != nil {
-		return fmt.Errorf("播放失败: %w", err)
-	}
-	return nil
-}
-
 func handleConnection(conn net.Conn, engine *TaskEngine) {
 	defer func() {
 		if r := recover(); r != nil {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@xdfnet/ispeak",
-  "version": "1.6.6",
+  "version": "1.6.7",
   "description": "Local macOS TTS daemon for AI coding assistants, powered by Volcengine streaming TTS.",
   "license": "MIT",
   "homepage": "https://github.com/xdfnet/iSpeak#readme",

package/scripts/ispeak

@@ -2,7 +2,7 @@
 # ispeak — iSpeak 控制命令
 set -euo pipefail
 
-VERSION="1.6.6"
+VERSION="1.6.7"
 SOCK="$HOME/.config/iSpeak/ispeak.sock"
 PLIST="$HOME/Library/LaunchAgents/com.iSpeak.plist"
 CMD_NAME="$(basename "$0")"