@goodandready/dsh-voice 0.8.17 → 0.8.18

package/lib/client.js

@@ -155,13 +155,6 @@ window.__ModuleLoader__.load({
       'sensevoiceBinHint': 'used when autostart is on',
       'sensevoiceModel': 'SenseVoice: model path',
       'sensevoiceAutostart': 'Autostart the SenseVoice server',
-      'realtimeStreaming': 'Realtime audio streaming',
-      'realtimeStreamingHint': 'Low-latency streaming via WebSocket (OpenAI Realtime API / sherpa-onnx)',
-      'realtimeProvider': 'Realtime provider',
-      'realtimeModel': 'Realtime model',
-      'realtimeActive': 'Streaming live…',
-      'realtimeConnect': 'Start streaming',
-      'realtimeDisconnect': 'Stop streaming',
       'visualizerStyle': 'Audio visualizer style',
       'visualizerStyleHint': 'Waveform animation inside the recording pill',
       'visLiquidWave': 'Liquid Wave',
@@ -302,13 +295,6 @@ window.__ModuleLoader__.load({
       'sensevoiceBinHint': 'используется при автозапуске',
       'sensevoiceModel': 'SenseVoice: путь к модели',
       'sensevoiceAutostart': 'Автозапуск сервера SenseVoice',
-      'realtimeStreaming': 'Потоковое аудио в реальном времени',
-      'realtimeStreamingHint': 'Низколатентный стриминг через WebSocket (OpenAI Realtime API / sherpa-onnx)',
-      'realtimeProvider': 'Провайдер реалтайма',
-      'realtimeModel': 'Модель реалтайма',
-      'realtimeActive': 'Стриминг идёт…',
-      'realtimeConnect': 'Начать стриминг',
-      'realtimeDisconnect': 'Остановить стриминг',
       'visualizerStyle': 'Стиль визуализатора звука',
       'visualizerStyleHint': 'Анимация волны внутри полоски записи',
       'visLiquidWave': 'Жидкая волна (Liquid Wave)',
@@ -529,11 +515,12 @@ window.__ModuleLoader__.load({
     // Как клавиша называется на человеческом языке.
     const KEY_LABELS = {
       Control: 'Ctrl', Alt: 'Alt', Shift: 'Shift', Meta: 'Win',
-      Space: t('keySpace'), Escape: 'Esc',
+      Escape: 'Esc',
     }
 
     function keyLabel(name) {
       if (!name) return t('keyUnset')
+      if (name === 'Space') return t('keySpace')
       if (KEY_LABELS[name]) return KEY_LABELS[name]
       // Коды вида KeyR и Digit5 показываем без служебной приставки.
       return String(name).replace(/^Key/, '').replace(/^Digit/, '')
@@ -777,7 +764,14 @@ window.__ModuleLoader__.load({
               voice.browser = null
               voice.caption = ''
               openMic(mode)
-                .then((rec) => { voice.rec = rec; voice.notify() })
+                .then((rec) => {
+                  if (voice.phase !== 'recording') {
+                    teardown(rec)
+                    return
+                  }
+                  voice.rec = rec
+                  voice.notify()
+                })
                 .catch((err) => voice.set({ phase: 'error', error: String(err && err.message ? err.message : err), rec: null }))
             }
           }
@@ -811,7 +805,14 @@ window.__ModuleLoader__.load({
         // Браузерная нога — только если её прямо поставили первой в цепочке.
         if (info.chain[0] === 'browser' && startBrowserLeg(mode, info.language)) return
         openMic(mode)
-          .then((rec) => { voice.rec = rec; voice.notify() })
+          .then((rec) => {
+            if (voice.phase !== 'recording') {
+              teardown(rec)
+              return
+            }
+            voice.rec = rec
+            voice.notify()
+          })
           .catch((err) => voice.set({ phase: 'error', error: String(err && err.message ? err.message : err), rec: null }))
       })
     }
@@ -1971,24 +1972,6 @@ window.__ModuleLoader__.load({
               onChange: (e) => setTop('sensevoiceAutostart', e.target.checked),
             })),
         ),
-        React.createElement('div', { className: 'dvs-block' },
-          React.createElement('div', { className: 'dvs-h' }, t('realtimeStreaming')),
-          React.createElement('div', { className: 'dvs-sub' }, t('realtimeStreamingHint')),
-          React.createElement('label', { className: 'dvs-field' }, t('realtimeStreaming'),
-            React.createElement('input', {
-              type: 'checkbox', checked: !!(draft && draft.realtimeStreaming), disabled: !writable,
-              onChange: (e) => setTop('realtimeStreaming', e.target.checked),
-            })),
-          React.createElement('label', { className: 'dvs-field' }, t('realtimeProvider'),
-            React.createElement('select', {
-              value: String((draft && draft.realtimeProvider) || 'openai'), disabled: !writable,
-              onChange: (e) => setTop('realtimeProvider', e.target.value),
-            },
-              React.createElement('option', { value: 'openai' }, 'OpenAI Realtime API'),
-              React.createElement('option', { value: 'sherpa-onnx' }, 'Sherpa-ONNX (local)'),
-            )),
-          textField('realtimeModel', t('realtimeModel'), ''),
-        ),
         React.createElement('div', { className: 'dvs-block' },
           React.createElement('div', { className: 'dvs-h' }, t('providerDashboard')),
           React.createElement('div', { className: 'dvo-dash' },

package/lib/index.js

@@ -155,12 +155,6 @@ export const Config = z.object({
     .description('SenseVoice / sherpa-onnx model path or model identifier.'),
   sensevoiceAutostart: z.boolean().default(false)
     .description('Start the local SenseVoice / sherpa-onnx server process automatically on startup.'),
-  realtimeStreaming: z.boolean().default(false)
-    .description('Enable low-latency streaming recognition (OpenAI Realtime API or Sherpa-ONNX).'),
-  realtimeProvider: z.string().default('openai')
-    .description('Realtime streaming provider: "openai" (OpenAI Realtime API) or "sherpa-onnx" (local streaming).'),
-  realtimeModel: z.string().default('gpt-4o-realtime-preview')
-    .description('Realtime model identifier.'),
   visualizerStyle: z.union(['liquid-wave', 'dynamic-orb', 'bars', 'off']).default('liquid-wave')
     .description('Audio visualizer animation style in the recording pill: "liquid-wave", "dynamic-orb", "bars", or "off".'),
 })
@@ -442,9 +436,6 @@ export function apply(ctx, baseConfig) {
         sensevoiceUrl: cfg.sensevoiceUrl,
         sensevoiceAutostart: !!cfg.sensevoiceAutostart,
         deepgramBaseUrl: cfg.deepgramBaseUrl || 'https://api.deepgram.com',
-        realtimeStreaming: !!cfg.realtimeStreaming,
-        realtimeProvider: cfg.realtimeProvider || 'openai',
-        realtimeModel: cfg.realtimeModel || 'gpt-4o-realtime-preview',
         visualizerStyle: cfg.visualizerStyle || 'liquid-wave',
         providerStats: statsTracker.get(),
       })
@@ -586,226 +577,8 @@ export function apply(ctx, baseConfig) {
     }),
   )
 
-  // ─── Realtime WebSocket Bridge ───────────────────────────────────────
-  // Клиент открывает WS на /dsh-voice/realtime, шлёт бинарные аудио-чанки.
-  // Хост открывает WS к провайдеру (OpenAI Realtime API или совместимому),
-  // пересылает аудио и получает текстовые дельты обратно клиенту.
-  // API-ключи никогда не покидают хост.
-  const realtimeSessions = new Set()
-
-  ctx.effect(() => ctx.webServer.register({
-    kind: 'exact',
-    path: '/dsh-voice/realtime',
-    handler: async (req, res) => {
-      const cfg = live()
-      if (!cfg.realtimeStreaming) {
-        writeJson(res, 403, { ok: false, error: { code: 'disabled', message: 'realtime streaming is disabled' } })
-        return
-      }
-      // Только WebSocket upgrade
-      if (!req.headers.upgrade || req.headers.upgrade.toLowerCase() !== 'websocket') {
-        writeJson(res, 426, { ok: false, error: { code: 'upgrade', message: 'WebSocket upgrade required' } })
-        return
-      }
-
-      // Получаем API-ключ для upstream-провайдера
-      const provider = cfg.realtimeProvider || 'openai'
-      let apiKey = ''
-      if (provider === 'openai') {
-        apiKey = await resolveKey('OPENAI_API_KEY')
-      }
-
-      // Определяем upstream WebSocket URL
-      let upstreamUrl
-      const model = cfg.realtimeModel || 'gpt-4o-realtime-preview'
-      if (provider === 'openai') {
-        upstreamUrl = `wss://api.openai.com/v1/realtime?model=${encodeURIComponent(model)}`
-      } else if (provider === 'sherpa-onnx') {
-        // Локальный sherpa-onnx WebSocket endpoint
-        const base = String(cfg.sensevoiceUrl || '').replace(/^http/, 'ws').replace(/\/api\/.*$/, '').replace(/\/v1\/.*$/, '')
-        upstreamUrl = base + '/ws/transcribe'
-      } else {
-        writeJson(res, 400, { ok: false, error: { code: 'provider', message: `unknown realtime provider: ${provider}` } })
-        return
-      }
-
-      // Upgrade клиентский запрос в WebSocket через ручной handshake
-      // Node.js http.Server не имеет встроенного WS — используем raw socket
-      const crypto = await import('node:crypto')
-      const wsKey = req.headers['sec-websocket-key']
-      if (!wsKey) {
-        res.writeHead(400); res.end('Missing Sec-WebSocket-Key'); return
-      }
-      const accept = crypto.createHash('sha1')
-        .update(wsKey + '258EAFA5-E914-47DA-95CA-5AB5ADF35F20')
-        .digest('base64')
-
-      res.writeHead(101, {
-        Upgrade: 'websocket',
-        Connection: 'Upgrade',
-        'Sec-WebSocket-Accept': accept,
-      })
-
-      const clientSocket = req.socket
-
-      // Простой WS frame parser/writer для бинарных и текстовых сообщений
-      function writeWsFrame(socket, data, opcode = 0x01) {
-        const payload = typeof data === 'string' ? Buffer.from(data, 'utf8') : data
-        const len = payload.length
-        let header
-        if (len < 126) {
-          header = Buffer.alloc(2)
-          header[0] = 0x80 | opcode
-          header[1] = len
-        } else if (len < 65536) {
-          header = Buffer.alloc(4)
-          header[0] = 0x80 | opcode
-          header[1] = 126
-          header.writeUInt16BE(len, 2)
-        } else {
-          header = Buffer.alloc(10)
-          header[0] = 0x80 | opcode
-          header[1] = 127
-          header.writeBigUInt64BE(BigInt(len), 2)
-        }
-        socket.write(Buffer.concat([header, payload]))
-      }
-
-      // Открываем upstream WebSocket через глобальный WebSocket (Node 22+)
-      let upstream = null
-      try {
-        const headers = {}
-        if (apiKey) headers['Authorization'] = `Bearer ${apiKey}`
-        if (provider === 'openai') headers['OpenAI-Beta'] = 'realtime=v1'
-        upstream = new WebSocket(upstreamUrl, { headers })
-      } catch (e) {
-        writeWsFrame(clientSocket, JSON.stringify({ type: 'error', message: String(e && e.message || e) }))
-        clientSocket.end()
-        return
-      }
-
-      const sessionId = crypto.randomUUID()
-      realtimeSessions.add(sessionId)
-
-      upstream.addEventListener('open', () => {
-        writeWsFrame(clientSocket, JSON.stringify({ type: 'session.created', sessionId }))
-        // Для OpenAI Realtime — отправить session.update с конфигурацией
-        if (provider === 'openai') {
-          upstream.send(JSON.stringify({
-            type: 'session.update',
-            session: {
-              modalities: ['text'],
-              input_audio_format: 'pcm16',
-              input_audio_transcription: { model: 'whisper-1' },
-              turn_detection: { type: 'server_vad' },
-            },
-          }))
-        }
-      })
-
-      upstream.addEventListener('message', (event) => {
-        // Пересылаем текстовые сообщения от upstream к клиенту
-        const msg = typeof event.data === 'string' ? event.data : event.data.toString()
-        try {
-          const parsed = JSON.parse(msg)
-          // Фильтруем и пересылаем только полезные события
-          if (parsed.type === 'conversation.item.input_audio_transcription.completed'
-            || parsed.type === 'response.audio_transcript.delta'
-            || parsed.type === 'response.audio_transcript.done'
-            || parsed.type === 'input_audio_buffer.speech_started'
-            || parsed.type === 'input_audio_buffer.speech_stopped'
-            || parsed.type === 'error') {
-            writeWsFrame(clientSocket, msg)
-          }
-        } catch {
-          // Не JSON — пересылаем как есть
-          writeWsFrame(clientSocket, msg)
-        }
-      })
-
-      upstream.addEventListener('error', () => {
-        writeWsFrame(clientSocket, JSON.stringify({ type: 'error', message: 'upstream connection error' }))
-      })
-
-      upstream.addEventListener('close', () => {
-        writeWsFrame(clientSocket, JSON.stringify({ type: 'session.closed' }))
-        try { clientSocket.end() } catch { /* socket уже закрыт */ }
-        realtimeSessions.delete(sessionId)
-      })
-
-      // Парсим входящие WS фреймы от клиента
-      let frameBuf = Buffer.alloc(0)
-      clientSocket.on('data', (chunk) => {
-        frameBuf = Buffer.concat([frameBuf, chunk])
-        while (frameBuf.length >= 2) {
-          const opcode = frameBuf[0] & 0x0f
-          const masked = !!(frameBuf[1] & 0x80)
-          let payloadLen = frameBuf[1] & 0x7f
-          let offset = 2
-          if (payloadLen === 126) {
-            if (frameBuf.length < 4) return
-            payloadLen = frameBuf.readUInt16BE(2)
-            offset = 4
-          } else if (payloadLen === 127) {
-            if (frameBuf.length < 10) return
-            payloadLen = Number(frameBuf.readBigUInt64BE(2))
-            offset = 10
-          }
-          if (masked) offset += 4
-          if (frameBuf.length < offset + payloadLen) return
-
-          let payload = frameBuf.subarray(offset, offset + payloadLen)
-          if (masked) {
-            const mask = frameBuf.subarray(offset - 4, offset)
-            payload = Buffer.from(payload)
-            for (let i = 0; i < payload.length; i++) payload[i] ^= mask[i & 3]
-          }
-          frameBuf = frameBuf.subarray(offset + payloadLen)
-
-          if (opcode === 0x08) {
-            // Close frame
-            upstream.close()
-            try { clientSocket.end() } catch { /* ok */ }
-            realtimeSessions.delete(sessionId)
-            return
-          }
-          if (opcode === 0x09) {
-            // Ping → Pong
-            writeWsFrame(clientSocket, payload, 0x0a)
-            continue
-          }
-          if (opcode === 0x01) {
-            // Текст: JSON команда от клиента → upstream
-            try { upstream.send(payload.toString('utf8')) } catch { /* upstream закрыт */ }
-          } else if (opcode === 0x02) {
-            // Бинарные данные: аудио-чанк → upstream как input_audio_buffer.append
-            if (provider === 'openai') {
-              upstream.send(JSON.stringify({
-                type: 'input_audio_buffer.append',
-                audio: payload.toString('base64'),
-              }))
-            } else {
-              // Для sherpa-onnx — отправляем бинарные данные напрямую
-              try { upstream.send(payload) } catch { /* upstream закрыт */ }
-            }
-          }
-        }
-      })
-
-      clientSocket.on('close', () => {
-        try { upstream.close() } catch { /* ok */ }
-        realtimeSessions.delete(sessionId)
-      })
-      clientSocket.on('error', () => {
-        try { upstream.close() } catch { /* ok */ }
-        realtimeSessions.delete(sessionId)
-      })
-    },
-  }), 'dsh-voice: /realtime route')
-
   ctx.effect(() => () => {
     if (child) { try { child.kill && child.kill() } catch { /* уже мёртв */ } }
     if (sensevoiceChild) { try { sensevoiceChild.kill && sensevoiceChild.kill() } catch { /* уже мёртв */ } }
-    for (const sid of realtimeSessions) realtimeSessions.delete(sid)
   }, 'dsh-voice: stop child processes')
 }

package/lib/normalize.js

@@ -29,7 +29,7 @@ const NUM_WORDS = {
 }
 
 const NUM_NAMES = Object.keys(NUM_WORDS).sort((a, b) => b.length - a.length).join('|')
-const NUM_SEQ = new RegExp('(' + NUM_NAMES + ')(?:\\s+(' + NUM_NAMES + '))*', 'gi')
+const NUM_SEQ = new RegExp('(?<![а-яёa-z0-9])(?:' + NUM_NAMES + ')(?:\\s+(?:' + NUM_NAMES + '))*(?![а-яёa-z0-9])', 'gi')
 
 export function wordsToDigits(text) {
   return text.replace(NUM_SEQ, (m) => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@goodandready/dsh-voice",
-  "version": "0.8.17",
+  "version": "0.8.18",
   "description": "Voice input for DeepSeek Harness: dictation chunked by pauses and voice messages, each with its own provider fallback chain (Deepgram, Groq, HuggingFace, local whisper.cpp, plus any OpenAI-compatible API of your own).",
   "license": "MIT",
   "type": "module",