dsh-fish-pet 0.2.5 → 0.2.6

package/CHANGELOG.md

@@ -1,5 +1,15 @@
 # Changelog
 
+## [0.2.6] - 2026-10-31
+
+### Changed
+- 回复长度限制 60 字 → **30 字**（越短开口越快）
+- 常驻 TTS 进程：单句合成 4 秒 → **2 秒**
+- 分段流水线：先合成第一句就开播，后面边播边合成，开口等待再减半
+- 中英历史隔离：中文模式只带中文历史、英文模式只带英文历史，避免串味
+
+### Fixed
+- 中文提问偶尔用英文回答（历史残留英文带跑）
 ## [0.2.5] - 2026-10-31
 
 ### Added

package/README.md

@@ -139,34 +139,50 @@ powershell -ExecutionPolicy Bypass -File tools\encrypt-keys.ps1
 
 之后鱼和耳朵会优先读 `deepseek.key.enc` / `siliconflow.key.enc`，明文文件会被自动删除。
 
-## 配置文件
-
-首次启动会生成 `%USERPROFILE%\.dsh\fish-pet\config.json`，可以改：
-
-```json
-{
-  "persona": {
-    "name": "大肥鱼",
-    "dad": "光光",
-    "mom": "丽丽",
-    "birthday": "2026年10月30日",
-    "birthplace": "中国广西北海",
-    "likes": "白饭、被摸头、主人夸它",
-    "dislikes": "被说胖、被冷落、吵",
-    "extra": "你说话简短、俏皮、微微傲娇，用中文，称呼用户为\"主人\"，每次回复不超过60个字，不要用表情符号。"
-  },
-  "bubble": { "opacity": 0.15 },
-  "voice": { "silenceGate": 1200, "boost": 1, "engine": "cloud", "localConf": 0.55 },
-  "tts": { "engine": "edge", "voice": "zh-CN-XiaoxiaoNeural", "rate": "+12%", "pitch": "+10Hz" }
-}
-```
-
-- `persona`：名字、家人、生日、出生地、喜好、厌恶、说话风格——改完重启鱼即生效
-- `bubble.opacity`：气泡文字透明度（0 = 全透明，1 = 不透明）
-- `voice.silenceGate`：说话要多响才算"开始说话"（越大越不灵敏）
-- `voice.boost`：麦克风增益倍数
-- `voice.engine`：语音识别引擎，`cloud`（硅基流动 Qwen3-ASR，最准）/ `local`（Windows 自带，免费但吃字）/ `auto`（本地优先、云端兜底）
-- `voice.localConf`：`auto` 模式下本地识别置信度门槛，低于它就转云端
+## 配置文件
+
+首次启动会生成 `%USERPROFILE%\.dsh\fish-pet\config.json`，可以改：
+
+```json
+{
+  "persona": {
+    "name": "大肥鱼",
+    "nameEn": "Mia",
+    "dad": "光光",
+    "mom": "丽丽",
+    "birthday": "2026年10月30日",
+    "birthplace": "中国广西北海",
+    "likes": "白饭、被摸头、主人夸它",
+    "dislikes": "被说胖、被冷落、吵",
+    "lang": "auto",
+    "extra": "你说话简短、俏皮、微微傲娇，称呼用户为\"主人\"，每次回复不超过30个字，不要用表情符号。"
+  },
+  "bubble": { "opacity": 0.15 },
+  "voice": { "silenceGate": 600, "silenceMs": 900, "boost": 1, "engine": "cloud", "localConf": 0.55 },
+  "tts": {
+    "engine": "edge",
+    "voice": "zh-CN-XiaoxiaoNeural",
+    "rate": "+12%",
+    "pitch": "+10Hz",
+    "voiceEn": "en-US-AnaNeural",
+    "rateEn": "+8%",
+    "pitchEn": "+15Hz"
+  }
+}
+```
+
+- `persona.name`：中文名（说中文时自称）
+- `persona.nameEn`：**英文名**（说英文时自称，默认 `Mia`／米娅）
+- `persona.dad` / `mom` / `birthday` / `birthplace`：家人、生日、出生地
+- `persona.likes` / `dislikes`：喜好、厌恶
+- `persona.lang`：语言模式，`zh`（只用中文）/ `en`（只用英文）/ `auto`（跟着你说，默认）
+- `persona.extra`：说话风格与长度限制（默认 30 字以内，回复越短开口越快）
+- `bubble.opacity`：气泡文字透明度（0 = 全透明，1 = 不透明）
+- `voice.silenceGate`：说话要多响才算"开始说话"（越小越灵敏，默认 600）
+- `voice.silenceMs`：说完后等多久算"讲完了"（英文建议 900~1100，太小会把句子切断）
+- `voice.boost`：麦克风增益倍数
+- `voice.engine`：语音识别引擎，`cloud`（硅基流动 Qwen3-ASR，最准）/ `local`（Windows 自带，免费但吃字）/ `auto`（本地优先、云端兜底）
+- `voice.localConf`：`auto` 模式下本地识别置信度门槛，低于它就转云端
 
 ### 语音朗读（TTS）
 

package/assets/asr-listener.ps1

@@ -2,7 +2,7 @@
 # 一次连续采集，检测到"说完"立刻上传，不再固定切片
 param(
     [int]$MaxMs = 10000,
-    [int]$SilenceMs = 1100,
+    [int]$SilenceMs = 900,
     [int]$Boost = 1,
     [int]$SilenceGate = 600
 )

package/assets/fish-pet.ps1

@@ -49,7 +49,7 @@ $StatePath = Join-Path $DataDir 'state.json'
 $ConfigPath = Join-Path $DataDir 'config.json'
 function Load-Config {
     $def = [ordered]@{
-        persona = [ordered]@{ name = '大肥鱼'; nameEn = 'Mia'; dad = '光光'; mom = '丽丽'; birthday = '2026年10月30日'; birthplace = '中国广西北海'; likes = '白饭、被摸头、主人夸它'; dislikes = '被说胖、被冷落、吵'; ownerBirthday = ''; lang = 'auto'; extra = '你说话简短、俏皮、微微傲娇，用中文，称呼用户为"主人"，每次回复不超过60个字，不要用表情符号。' }
+        persona = [ordered]@{ name = '大肥鱼'; nameEn = 'Mia'; dad = '光光'; mom = '丽丽'; birthday = '2026年10月30日'; birthplace = '中国广西北海'; likes = '白饭、被摸头、主人夸它'; dislikes = '被说胖、被冷落、吵'; ownerBirthday = ''; lang = 'auto'; extra = '你说话简短、俏皮、微微傲娇，用中文，称呼用户为"主人"，每次回复不超过30个字，不要用表情符号。' }
         bubble  = [ordered]@{ opacity = 0.15 }
         voice   = [ordered]@{ silenceGate = 1200; boost = 1; engine = 'cloud'; localConf = 0.55 }
         tts     = [ordered]@{ engine = 'edge'; voice = 'zh-CN-XiaoxiaoNeural'; rate = '+12%'; pitch = '+10Hz'; voiceEn = 'en-US-AnaNeural'; rateEn = '+8%'; pitchEn = '+15Hz' }
@@ -1137,38 +1137,135 @@ function Play-TtsFile([string]$mp3, [string]$voice) {
         if ($null -eq $script:mp) {
             Add-Type -AssemblyName PresentationCore
             $script:mp = New-Object System.Windows.Media.MediaPlayer
-            $script:mp.Add_MediaEnded({ $script:speakEndAt = (Get-Date) })
+            $script:mp.Add_MediaEnded({ Complete-TtsSegment })
             $script:mp.Add_MediaOpened({
                 try { if ($script:mp.NaturalDuration.HasTimeSpan) { $script:speakEndAt = (Get-Date).AddSeconds($script:mp.NaturalDuration.TimeSpan.TotalSeconds + 0.8) } } catch { }
             })
         }
         $script:mp.Open([uri]$mp3)
-        # 用时长估算"说到什么时候"，避免 MediaEnded 提前误触发导致自己听见自己
-        $dur = 6.0
-        try { if ($script:mp.NaturalDuration.HasTimeSpan) { $dur = $script:mp.NaturalDuration.TimeSpan.TotalSeconds } } catch { }
-        if ($dur -lt 0.5) { $dur = [Math]::Max(3.0, $text.Length * 0.22) }
-        $script:speakEndAt = (Get-Date).AddSeconds($dur + 0.8)
         $script:mp.Play()
         Write-Err ('TTS-EDGE voice=' + $voice + ' file=' + (Get-Item $mp3).Length)
         try { Get-ChildItem $env:TEMP -Filter 'fish-tts-*.mp3' -ErrorAction SilentlyContinue | Where-Object { $_.LastWriteTime -lt (Get-Date).AddMinutes(-10) } | Remove-Item -Force -ErrorAction SilentlyContinue } catch { }
     } catch { Write-Err ('TTS-PLAY-ERR ' + $_.Exception.Message) }
 }
 
+# 把回复拆成 1~3 段，边合成边播，缩短"开口等待"
+function Split-TtsSegments([string]$text) {
+    if (-not $text) { return @() }
+    $parts = @($text -split '(?<=[。！？!?；;…\.\r\n])' | Where-Object { $_.Trim() })
+    if ($parts.Count -le 1) { return @($text) }
+    # 第一段尽量短（先开口），后面合并成最多 2 段
+    $first = $parts[0].Trim()
+    $rest = ''
+    for ($i = 1; $i -lt $parts.Count; $i++) { $rest = $rest + $parts[$i] }
+    $rest = $rest.Trim()
+    if (-not $rest) { return @($text) }
+    if ($rest.Length -le 26) { return ,@($first, $rest) }
+    # 剩余拆成两半
+    $mid = [Math]::Floor($rest.Length / 2)
+    $cut = $rest.LastIndexOfAny([char[]]@('，', ',', '；', '、', ' ', '。', '.'), $mid)
+    if ($cut -lt 8) { $cut = $mid }
+    $a = $rest.Substring(0, $cut + 1).Trim()
+    $b = $rest.Substring($cut + 1).Trim()
+    if (-not $b) { return ,@($first, $a) }
+    return ,@($first, $a, $b)
+}
+
+function Start-TtsPipeline([string]$text, [string]$voice, [string]$rate, [string]$pitch) {
+    $segs = Split-TtsSegments $text
+    $script:ttsSegs = @($segs)
+    $script:ttsSegVoice = $voice
+    $script:ttsSegRate = $rate
+    $script:ttsSegPitch = $pitch
+    $script:ttsSegFiles = @()
+    $script:ttsSegDone = @()
+    $script:ttsSegPlayed = @()
+    $script:ttsSegReq = @()
+    for ($i = 0; $i -lt $script:ttsSegs.Count; $i++) {
+        $script:ttsSegFiles += (Join-Path $env:TEMP ('fish-tts-' + [guid]::NewGuid().ToString('N') + '.mp3'))
+        $script:ttsSegDone += $false
+        $script:ttsSegPlayed += $false
+        $script:ttsSegReq += ''
+    }
+    $script:ttsSegPlaying = -1
+    Write-Err ('TTS-PIPE segs=' + $script:ttsSegs.Count + ' len=' + $text.Length)
+    $script:ttsTimer.Start()
+}
+
+function Request-TtsSegment([int]$idx) {
+    if ($idx -lt 0 -or $idx -ge $script:ttsSegs.Count) { return }
+    if ($script:ttsSegReq[$idx]) { return }
+    try {
+        $reqId = '{0:D2}{1}' -f $idx, ([guid]::NewGuid().ToString('N').Substring(0, 8))
+        $reqFile = Join-Path $env:TEMP ('fish-tts-req-' + $reqId + '.json')
+        $reqObj = @{
+            id = $reqId
+            text = $script:ttsSegs[$idx]
+            voice = $script:ttsSegVoice
+            rate = $script:ttsSegRate
+            pitch = $script:ttsSegPitch
+            out = $script:ttsSegFiles[$idx]
+            status = $script:ttsStatusFile
+        } | ConvertTo-Json -Compress
+        [IO.File]::WriteAllText($reqFile, $reqObj, (New-Object Text.UTF8Encoding($false)))
+        $script:ttsProc.StandardInput.WriteLine('@' + $reqFile)
+        $script:ttsProc.StandardInput.Flush()
+        $script:ttsSegReq[$idx] = $reqId
+        Write-Err ('TTS-REQ idx=' + $idx + ' id=' + $reqId)
+    } catch { Write-Err ('TTS-REQ-ERR ' + $_.Exception.Message) }
+}
+
+function Complete-TtsSegment {
+    try {
+        $cur = $script:ttsSegPlaying
+        if ($cur -ge 0 -and $cur -lt $script:ttsSegPlayed.Count) { $script:ttsSegPlayed[$cur] = $true }
+        $script:ttsSegPlaying = -1
+        $script:speakEndAt = (Get-Date)
+    } catch { }
+}
+
+# 启动常驻 TTS 进程（复用连接，比每次启动 python 快一倍）
+function Start-TtsWorker {
+    try {
+        if ($script:ttsProc -and -not $script:ttsProc.HasExited) { return $true }
+        $workerPath = Join-Path (Split-Path -Parent $PSCommandPath) 'tts_worker.py'
+        if (-not (Test-Path $workerPath)) { Write-Err 'TTS-WORKER-MISSING'; return $false }
+        $psi = New-Object System.Diagnostics.ProcessStartInfo
+        $psi.FileName = 'python'
+        $psi.Arguments = '"' + $workerPath + '"'
+        $psi.UseShellExecute = $false
+        $psi.RedirectStandardInput = $true
+        $psi.CreateNoWindow = $true
+        $script:ttsProc = [System.Diagnostics.Process]::Start($psi)
+        Start-Sleep -Milliseconds 350
+        if ($script:ttsProc.HasExited) { Write-Err 'TTS-WORKER-DIED'; return $false }
+        $script:ttsStatusFile = Join-Path $env:TEMP 'fish-tts-status.txt'
+        Write-Err ('TTS-WORKER-START pid=' + $script:ttsProc.Id)
+        return $true
+    } catch { Write-Err ('TTS-WORKER-ERR ' + $_.Exception.Message); return $false }
+}
+
 function Speak-Text([string]$text) {
     if (-not $script:speakOn) { return }
     $talkFile = Join-Path $DataDir 'asr.talking'
     try { [IO.File]::WriteAllText($talkFile, '1') } catch { }
     if ($script:ttsEngine -eq 'edge') {
         try {
-            $mp3 = Join-Path $env:TEMP ('fish-tts-' + [guid]::NewGuid().ToString('N') + '.mp3')
-            # 英文回复用英文音色，中文用中文音色
             $spkLang = Get-TextLang $text
             if ($spkLang -eq 'en') { $vUse = $script:ttsVoiceEn; $rUse = $script:ttsRateEn; $pUse = $script:ttsPitchEn } else { $vUse = $script:ttsVoice; $rUse = $script:ttsRate; $pUse = $script:ttsPitch }
-            # 后台线程生成音频：不阻塞界面和游动
+            if (Start-TtsWorker) {
+                $script:ttsMode = 'worker'
+                $script:ttsVoiceUsed = $vUse
+                $script:ttsText = $text
+                Start-TtsPipeline $text $vUse $rUse $pUse
+                return
+            }
+            # 常驻进程起不来：退回后台线程整段合成
             if ($null -eq $script:ttsPool) {
                 $script:ttsPool = [runspacefactory]::CreateRunspacePool(1, 2)
                 $script:ttsPool.Open()
             }
+            $mp3 = Join-Path $env:TEMP ('fish-tts-' + [guid]::NewGuid().ToString('N') + '.mp3')
             $ps = [PowerShell]::Create()
             $ps.RunspacePool = $script:ttsPool
             [void]$ps.AddScript({
@@ -1176,22 +1273,65 @@ function Speak-Text([string]$text) {
                 & $exe -m edge_tts --voice $voice ('--rate=' + $rate) ('--pitch=' + $pitch) '--text' $txt '--write-media' $out 2>$null | Out-Null
             }).AddArgument('python').AddArgument($vUse).AddArgument($rUse).AddArgument($pUse).AddArgument($text).AddArgument($mp3)
             $script:ttsPS = $ps
+            $script:ttsHandle = $ps.BeginInvoke()
+            $script:ttsMode = 'thread'
             $script:ttsMp3 = $mp3
-            $script:ttsText = $text
             $script:ttsVoiceUsed = $vUse
-            $script:ttsHandle = $ps.BeginInvoke()
+            $script:ttsText = $text
             $script:ttsTimer.Start()
-            Write-Err ('TTS-EDGE-START len=' + $text.Length + ' voice=' + $vUse)
+            Write-Err ('TTS-EDGE-START len=' + $text.Length + ' voice=' + $vUse + ' mode=thread')
             return
         } catch { Write-Err ('TTS-EDGE-ERR ' + $_.Exception.Message) }
     }
     Speak-Sapi $text
 }
 
-# 后台 TTS 完成检测（UI 线程轮询，拿回来再播放）
+# 流水线调度：边合成边播
 $script:ttsTimer = New-Object System.Windows.Forms.Timer
-$script:ttsTimer.Interval = 100
+$script:ttsTimer.Interval = 90
 $script:ttsTimer.Add_Tick({
+    # ---- 常驻进程 + 分段流水线 ----
+    if ($script:ttsMode -eq 'worker' -and $script:ttsSegs.Count -gt 0) {
+        # 1) 扫状态文件，标记完成的分段
+        for ($i = 0; $i -lt $script:ttsSegs.Count; $i++) {
+            if ($script:ttsSegDone[$i] -or -not $script:ttsSegReq[$i]) { continue }
+            $rid = $script:ttsSegReq[$i]
+            try {
+                $hit = Get-Content $script:ttsStatusFile -ErrorAction SilentlyContinue | Where-Object { $_ -match ('^' + $rid + ' ') } | Select-Object -Last 1
+            } catch { $hit = $null }
+            if ($hit) {
+                if ($hit -like ('*' + $rid + ' OK*')) { $script:ttsSegDone[$i] = $true }
+                else { Write-Err ('TTS-SEG-FAIL idx=' + $i); $script:ttsSegDone[$i] = $true }
+            }
+        }
+        # 2) 没有在播 → 找下一段
+        if ($script:ttsSegPlaying -lt 0) {
+            $next = -1
+            for ($i = 0; $i -lt $script:ttsSegs.Count; $i++) { if (-not $script:ttsSegPlayed[$i]) { $next = $i; break } }
+            if ($next -lt 0) {
+                $script:ttsTimer.Stop()
+                $script:ttsSegs = @()
+                return
+            }
+            if ($script:ttsSegDone[$next]) {
+                $file = $script:ttsSegFiles[$next]
+                if ((Test-Path $file) -and (Get-Item $file).Length -gt 800) {
+                    $script:ttsSegPlaying = $next
+                    Play-TtsFile $file $script:ttsVoiceUsed
+                } else {
+                    $script:ttsSegPlayed[$next] = $true
+                }
+            } elseif (-not $script:ttsSegReq[$next]) {
+                Request-TtsSegment $next
+            }
+        } else {
+            # 3) 正在播：预取下一段
+            $nx = $script:ttsSegPlaying + 1
+            if ($nx -lt $script:ttsSegs.Count -and -not $script:ttsSegDone[$nx] -and -not $script:ttsSegReq[$nx]) { Request-TtsSegment $nx }
+        }
+        return
+    }
+    # ---- 后台线程整段模式 ----
     if ($null -eq $script:ttsHandle) { $script:ttsTimer.Stop(); return }
     if (-not $script:ttsHandle.IsCompleted) { return }
     $script:ttsTimer.Stop()
@@ -1382,7 +1522,7 @@ function Start-Chat([string]$text) {
     $replyLang = Get-ReplyLang $text
     if ($replyLang -eq 'en') {
         $sys = 'YOU MUST REPLY IN ENGLISH ONLY, NO MATTER WHAT. You are a round, chubby blue fish living on the master''s Windows desktop. Your name is "' + $pn.nameEn + '" (in Chinese it is written 大肥鱼, said as Da Fei Yu). Your dad is Guangguang, your mom is Lili, your birthday is October 30 2026, you were born in Beihai, Guangxi, China. Answer questions about family/birthday/birthplace with these facts and never invent them. In English your name is always this one, never any other name; if earlier messages called you something different, correct them politely.'
-        $sys = $sys + ' Reply ONLY in English, max 40 words, playful and a little tsundere, call the user "master", no emoji.'
+        $sys = $sys + ' Reply ONLY in English, max 20 words, playful and a little tsundere, call the user "master", no emoji.'
         if ($pn.likes) { $sys = $sys + ' You like: ' + $pn.likes + ' .' }
         if ($pn.dislikes) { $sys = $sys + ' You dislike: ' + $pn.dislikes + ' .' }
     } else {
@@ -1424,10 +1564,13 @@ function Start-Chat([string]$text) {
         if ($histEn.Count -gt 4) { $histEn = $histEn[($histEn.Count - 4)..($histEn.Count - 1)] }
         foreach ($m in $histEn) { $msgs += $m }
     } else {
-        foreach ($m in $script:chatHistory) { $msgs += $m }
+        # 中文模式只带中文历史，避免被上一段英文带跑
+        $histZh = @($script:chatHistory | Where-Object { (Get-TextLang ([string]$_.content)) -eq 'zh' })
+        if ($histZh.Count -gt 10) { $histZh = $histZh[($histZh.Count - 10)..($histZh.Count - 1)] }
+        foreach ($m in $histZh) { $msgs += $m }
     }
     $msgs += @{ role = 'user'; content = $text }
-    $body  = @{ model = 'deepseek-flash'; messages = $msgs; max_tokens = 300; temperature = 0.9; thinking = @{ type = 'disabled' } } | ConvertTo-Json -Depth 6
+    $body  = @{ model = 'deepseek-flash'; messages = $msgs; max_tokens = 160; temperature = 0.9; thinking = @{ type = 'disabled' } } | ConvertTo-Json -Depth 6
     if ($null -eq $script:http) {
         $script:http = New-Object System.Net.Http.HttpClient
         $script:http.Timeout = [TimeSpan]::FromSeconds(40)

package/assets/tts_worker.py

@@ -42,10 +42,18 @@ async def main():
         line = line.strip()
         if not line:
             continue
-        try:
-            req = json.loads(line)
-        except Exception:
-            continue
+        # 支持"@文件路径"形式：避免 Windows PowerShell 写 stdin 时的编码问题
+        if line.startswith("@"):
+            try:
+                with open(line[1:], "r", encoding="utf-8") as fh:
+                    req = json.loads(fh.read())
+            except Exception:
+                continue
+        else:
+            try:
+                req = json.loads(line)
+            except Exception:
+                continue
         if req.get("cmd") == "quit":
             break
         try:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-fish-pet",
-  "version": "0.2.5",
+  "version": "0.2.6",
   "description": "Desktop blue fish pet for DeepSeek Harness: per-pixel transparent WinForms window, text chat, and cloud speech-to-text voice conversation.",
   "type": "module",
   "main": "lib/index.js",