iobroker.assistant 1.4.4 → 1.4.6

package/admin/i18n/pl.json

@@ -58,7 +58,7 @@
   "ESPHome satellites": "Satelity ESPHome",
   "ESPHome satellites do not get the reply over their API socket — they fetch it as a URL. This is the port that small HTTP server listens on.": "Satelity ESPHome nie odbierają odpowiedzi przez swoje gniazdo API — pobierają ją jako URL. To port, na którym nasłuchuje ten mały serwer HTTP.",
   "Empty = tts-1. Also: tts-1-hd, gpt-4o-mini-tts.": "Pusty = tts-1. Również: tts-1-hd, gpt-4o-mini-tts.",
-  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe.": "Pusty = szept-1. Również: gpt-4o-transcribe, gpt-4o-mini-transcribe.",
+  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe. With an own server: the model as that server names it, e.g. Systran/faster-whisper-large-v3-turbo.": "Puste = whisper-1. Również: gpt-4o-transcribe, gpt-4o-mini-transcribe. Z własnym serwerem: model tak, jak nazywa go serwer, np. Systran/faster-whisper-large-v3-turbo.",
   "Enable voice (STT/TTS for satellites)": "Włącz głos (STT/TTS dla satelitów)",
   "Enable voice server (satellites)": "Włącz serwer głosowy (satelity)",
   "Enables speech recognition/synthesis. ioBroker-native satellites reach it over the message bus — no port needed. Turn on the UDP server below only for ESP / Hannah satellites.": "Umożliwia rozpoznawanie/syntezę mowy. Satelity natywne dla ioBroker osiągają go przez magistralę komunikatów — nie jest potrzebny żaden port. Włącz serwer UDP poniżej tylko dla satelitów ESP/Hannah.",
@@ -222,5 +222,9 @@
   "Person": "Osoba",
   "Guest": "Gość",
   "Pet": "Zwierzę",
-  "Group": "Grupa"
+  "Group": "Grupa",
+  "Speech recognition server (optional)": "Serwer rozpoznawania mowy (opcjonalnie)",
+  "OpenAI-compatible server instead of OpenAI, e.g. a local Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Empty = OpenAI.": "Serwer zgodny z OpenAI zamiast OpenAI, np. lokalny Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Puste = OpenAI.",
+  "Key of the speech recognition server": "Klucz serwera rozpoznawania mowy",
+  "Only if the server wants one. The OpenAI key is never sent to it.": "Tylko jeśli serwer go wymaga. Klucz OpenAI nigdy nie jest tam wysyłany."
 }

package/admin/i18n/pt.json

@@ -58,7 +58,7 @@
   "ESPHome satellites": "Satélites ESPHome",
   "ESPHome satellites do not get the reply over their API socket — they fetch it as a URL. This is the port that small HTTP server listens on.": "Os satélites ESPHome não recebem a resposta pelo seu socket de API — buscam-na como URL. Esta é a porta em que esse pequeno servidor HTTP escuta.",
   "Empty = tts-1. Also: tts-1-hd, gpt-4o-mini-tts.": "Vazio = tts-1. Também: tts-1-hd, gpt-4o-mini-tts.",
-  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe.": "Vazio = whisper-1. Também: gpt-4o-transcribe, gpt-4o-mini-transcribe.",
+  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe. With an own server: the model as that server names it, e.g. Systran/faster-whisper-large-v3-turbo.": "Vazio = whisper-1. Também: gpt-4o-transcribe, gpt-4o-mini-transcribe. Com servidor próprio: o modelo como o servidor o chama, p. ex. Systran/faster-whisper-large-v3-turbo.",
   "Enable voice (STT/TTS for satellites)": "Ativar voz (STT/TTS para satélites)",
   "Enable voice server (satellites)": "Ativar servidor de voz (satélites)",
   "Enables speech recognition/synthesis. ioBroker-native satellites reach it over the message bus — no port needed. Turn on the UDP server below only for ESP / Hannah satellites.": "Habilita o reconhecimento/síntese de voz. Os satélites nativos do ioBroker acessam o servidor através do barramento de mensagens — sem necessidade de porta. Ative o servidor UDP abaixo apenas para satélites ESP/Hannah.",
@@ -222,5 +222,9 @@
   "Person": "Pessoa",
   "Guest": "Convidado",
   "Pet": "Animal de estimação",
-  "Group": "Grupo"
+  "Group": "Grupo",
+  "Speech recognition server (optional)": "Servidor de reconhecimento de voz (opcional)",
+  "OpenAI-compatible server instead of OpenAI, e.g. a local Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Empty = OpenAI.": "Servidor compatível com OpenAI em vez da OpenAI, p. ex. um Whisper local: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Vazio = OpenAI.",
+  "Key of the speech recognition server": "Chave do servidor de reconhecimento de voz",
+  "Only if the server wants one. The OpenAI key is never sent to it.": "Só se o servidor pedir uma. A chave da OpenAI nunca é enviada para lá."
 }

package/admin/i18n/ru.json

@@ -58,7 +58,7 @@
   "ESPHome satellites": "Сателлиты ESPHome",
   "ESPHome satellites do not get the reply over their API socket — they fetch it as a URL. This is the port that small HTTP server listens on.": "Сателлиты ESPHome получают ответ не через свой сокет API — они забирают его по URL. Это порт, на котором слушает этот небольшой HTTP-сервер.",
   "Empty = tts-1. Also: tts-1-hd, gpt-4o-mini-tts.": "Пусто = tts-1. Также: tts-1-hd, gpt-4o-mini-tts.",
-  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe.": "Пусто = whisper-1. Также: gpt-4o-transcribe, gpt-4o-mini-transcribe.",
+  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe. With an own server: the model as that server names it, e.g. Systran/faster-whisper-large-v3-turbo.": "Пусто = whisper-1. Также: gpt-4o-transcribe, gpt-4o-mini-transcribe. Со своим сервером: модель так, как её называет сервер, напр. Systran/faster-whisper-large-v3-turbo.",
   "Enable voice (STT/TTS for satellites)": "Включить голосовую связь (STT/TTS для спутниковой связи)",
   "Enable voice server (satellites)": "Включить голосовой сервер (спутники)",
   "Enables speech recognition/synthesis. ioBroker-native satellites reach it over the message bus — no port needed. Turn on the UDP server below only for ESP / Hannah satellites.": "Включает распознавание/синтез речи. Спутники, использующие ioBroker, получают к нему доступ через шину сообщений — порт не требуется. Включите UDP-сервер, указанный ниже, только для спутников ESP/Hannah.",
@@ -222,5 +222,9 @@
   "Person": "Человек",
   "Guest": "Гость",
   "Pet": "Питомец",
-  "Group": "Группа"
+  "Group": "Группа",
+  "Speech recognition server (optional)": "Сервер распознавания речи (опционально)",
+  "OpenAI-compatible server instead of OpenAI, e.g. a local Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Empty = OpenAI.": "OpenAI-совместимый сервер вместо OpenAI, напр. локальный Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Пусто = OpenAI.",
+  "Key of the speech recognition server": "Ключ сервера распознавания речи",
+  "Only if the server wants one. The OpenAI key is never sent to it.": "Только если сервер его требует. Ключ OpenAI туда никогда не отправляется."
 }

package/admin/i18n/uk.json

@@ -58,7 +58,7 @@
   "ESPHome satellites": "Сателіти ESPHome",
   "ESPHome satellites do not get the reply over their API socket — they fetch it as a URL. This is the port that small HTTP server listens on.": "Сателіти ESPHome отримують відповідь не через свій сокет API — вони забирають її за URL. Це порт, на якому слухає цей невеликий HTTP-сервер.",
   "Empty = tts-1. Also: tts-1-hd, gpt-4o-mini-tts.": "Порожнє = tts-1. Також: tts-1-hd, gpt-4o-mini-tts.",
-  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe.": "Порожнє = шепіт-1. Також: gpt-4o-trancribe, gpt-4o-mini-trancribe.",
+  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe. With an own server: the model as that server names it, e.g. Systran/faster-whisper-large-v3-turbo.": "Порожньо = whisper-1. Також: gpt-4o-transcribe, gpt-4o-mini-transcribe. З власним сервером: модель так, як її називає сервер, напр. Systran/faster-whisper-large-v3-turbo.",
   "Enable voice (STT/TTS for satellites)": "Увімкнути голосовий зв'язок (STT/TTS для супутників)",
   "Enable voice server (satellites)": "Увімкнути голосовий сервер (супутники)",
   "Enables speech recognition/synthesis. ioBroker-native satellites reach it over the message bus — no port needed. Turn on the UDP server below only for ESP / Hannah satellites.": "Дозволяє розпізнавання/синтез мовлення. Супутники, сумісні з ioBroker, досягають його через шину повідомлень — порт не потрібен. Увімкніть UDP-сервер нижче лише для супутників ESP / Hannah.",
@@ -222,5 +222,9 @@
   "Person": "Людина",
   "Guest": "Гість",
   "Pet": "Улюбленець",
-  "Group": "Група"
+  "Group": "Група",
+  "Speech recognition server (optional)": "Сервер розпізнавання мовлення (необов'язково)",
+  "OpenAI-compatible server instead of OpenAI, e.g. a local Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Empty = OpenAI.": "OpenAI-сумісний сервер замість OpenAI, напр. локальний Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Порожньо = OpenAI.",
+  "Key of the speech recognition server": "Ключ сервера розпізнавання мовлення",
+  "Only if the server wants one. The OpenAI key is never sent to it.": "Лише якщо сервер його вимагає. Ключ OpenAI туди ніколи не надсилається."
 }

package/admin/i18n/zh-cn.json

@@ -58,7 +58,7 @@
   "ESPHome satellites": "ESPHome 卫星",
   "ESPHome satellites do not get the reply over their API socket — they fetch it as a URL. This is the port that small HTTP server listens on.": "ESPHome 卫星不会通过其 API 套接字接收回复，而是按 URL 获取。这里就是那个小型 HTTP 服务器监听的端口。",
   "Empty = tts-1. Also: tts-1-hd, gpt-4o-mini-tts.": "空 = tts-1。另见：tts-1-hd、gpt-4o-mini-tts。",
-  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe.": "空 = whisper-1。另见：gpt-4o-transcribe、gpt-4o-mini-transcribe。",
+  "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe. With an own server: the model as that server names it, e.g. Systran/faster-whisper-large-v3-turbo.": "空 = whisper-1。另见：gpt-4o-transcribe、gpt-4o-mini-transcribe。使用自己的服务器时：按服务器的名称填写模型，例如 Systran/faster-whisper-large-v3-turbo。",
   "Enable voice (STT/TTS for satellites)": "启用语音功能（卫星语音转语音/文本转语音）",
   "Enable voice server (satellites)": "启用语音服务器（卫星）",
   "Enables speech recognition/synthesis. ioBroker-native satellites reach it over the message bus — no port needed. Turn on the UDP server below only for ESP / Hannah satellites.": "启用语音识别/合成功能。ioBroker 原生卫星通过消息总线与其通信，无需端口。仅当 ESP/Hannah 卫星连接时，才需要启用下方的 UDP 服务器。",
@@ -222,5 +222,9 @@
   "Person": "人员",
   "Guest": "访客",
   "Pet": "宠物",
-  "Group": "组"
+  "Group": "组",
+  "Speech recognition server (optional)": "语音识别服务器（可选）",
+  "OpenAI-compatible server instead of OpenAI, e.g. a local Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Empty = OpenAI.": "代替 OpenAI 的 OpenAI 兼容服务器，例如本地 Whisper：http://192.168.1.10:8000/v1（speaches、faster-whisper-server、LocalAI）。空 = OpenAI。",
+  "Key of the speech recognition server": "语音识别服务器的密钥",
+  "Only if the server wants one. The OpenAI key is never sent to it.": "仅当服务器需要时填写。OpenAI 密钥永远不会发送到该服务器。"
 }

package/admin/jsonConfig.json

@@ -1184,8 +1184,30 @@
                     "newLine": true,
                     "type": "text",
                     "label": "OpenAI STT model (optional)",
-                    "help": "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe.",
-                    "hidden": "!data.voiceEnabled || data.sttProvider !== 'openai'",
+                    "help": "Empty = whisper-1. Also: gpt-4o-transcribe, gpt-4o-mini-transcribe. With an own server: the model as that server names it, e.g. Systran/faster-whisper-large-v3-turbo.",
+                    "hidden": "!data.voiceEnabled || (data.sttProvider !== 'openai' && data.sttFallback !== 'openai')",
+                    "xs": 12,
+                    "sm": 6,
+                    "md": 4,
+                    "lg": 4,
+                    "xl": 4
+                },
+                "sttBaseUrl": {
+                    "type": "text",
+                    "label": "Speech recognition server (optional)",
+                    "help": "OpenAI-compatible server instead of OpenAI, e.g. a local Whisper: http://192.168.1.10:8000/v1 (speaches, faster-whisper-server, LocalAI). Empty = OpenAI.",
+                    "hidden": "!data.voiceEnabled || (data.sttProvider !== 'openai' && data.sttFallback !== 'openai')",
+                    "xs": 12,
+                    "sm": 6,
+                    "md": 4,
+                    "lg": 4,
+                    "xl": 4
+                },
+                "sttServerKey": {
+                    "type": "password",
+                    "label": "Key of the speech recognition server",
+                    "help": "Only if the server wants one. The OpenAI key is never sent to it.",
+                    "hidden": "!data.voiceEnabled || (data.sttProvider !== 'openai' && data.sttFallback !== 'openai') || !data.sttBaseUrl",
                     "xs": 12,
                     "sm": 6,
                     "md": 4,

package/build/lib/voice/engines.js

@@ -104,6 +104,9 @@ function buildSttEngine(provider, ctx) {
         case 'piper':
             throw new Error('Piper is a text-to-speech engine, not speech-to-text');
         default:
+            if (ctx.sttBaseUrl?.trim()) {
+                return new stt_1.OpenAiStt(ctx.sttServerKey || '', ctx.sttModel || undefined, ctx.sttBaseUrl);
+            }
             requireOpenAi(ctx.creds);
             return new stt_1.OpenAiStt(ctx.creds.openaiKey, ctx.sttModel || undefined);
     }

package/build/lib/voice/engines.js.map

@@ -1 +1 @@
-{"version":3,"file":"engines.js","sourceRoot":"","sources":["../../../src/lib/voice/engines.ts"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AA+FA,0CAYC;AAuBD,0CAeC;AAuDD,gCAmBC;AAGD,sCAKC;AAGD,sCAEC;AAxOD;;;;GAIG;AACH,+BAAkD;AAClD,+BAAoE;AACpE,mCAA8D;AAC9D,+BAAuE;AACvE,iCAAiD;AACjD,mCAAoD;AACpD,6CAOsB;AACtB,gDAAkC;AAElC,yCAAuC;AACvC,yCAAwE;AACxE,iCAAqC;AAgDrC,SAAS,YAAY,CAAC,KAAuB;IACzC,IAAI,CAAC,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,CAAC,WAAW,EAAE,CAAC;QACxC,MAAM,IAAI,KAAK,CAAC,kDAAkD,CAAC,CAAC;IACxE,CAAC;AACL,CAAC;AAED,SAAS,UAAU,CAAC,KAAuB;IACvC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,WAAW,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,eAAe,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,MAAM,EAAE,CAAC;QAC5E,MAAM,IAAI,KAAK,CAAC,iEAAiE,CAAC,CAAC;IACvF,CAAC;AACL,CAAC;AAED,SAAS,aAAa,CAAC,KAAuB;IAC1C,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,CAAC;QACnB,MAAM,IAAI,KAAK,CAAC,gCAAgC,CAAC,CAAC;IACtD,CAAC;AACL,CAAC;AAED,SAAS,iBAAiB,CAAC,KAAuB;IAC9C,IAAI,CAAC,KAAK,CAAC,aAAa,EAAE,CAAC;QACvB,MAAM,IAAI,KAAK,CAAC,oCAAoC,CAAC,CAAC;IAC1D,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAClG,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IAC1G,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC;QACnE,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QACrC,KAAK,MAAM;YACP,OAAO,IAAI,cAAO,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,SAAS,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAC5D,KAAK,YAAY;YACb,iBAAiB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YAC7B,OAAO,IAAI,0BAAa,CAAC,GAAG,CAAC,KAAK,CAAC,aAAa,EAAE,GAAG,CAAC,kBAAkB,CAAC,CAAC;QAC9E,KAAK,OAAO;YACR,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QAC5E;YACI,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAAC,GAAG,CAAC,KAAK,CAAC,SAAS,EAAE,GAAG,CAAC,QAAQ,IAAI,SAAS,CAAC,CAAC;IAC7E,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,sGAAsG;IACtG,iGAAiG;IACjG,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAC7F,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IACrG,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,sGAAsG;QACtG,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED;;;GAGG;AACH,SAAS,SAAS,CAAC,QAAwB,EAAE,GAAkB;IAC3D,MAAM,KAAK,GACP,QAAQ,KAAK,OAAO;QAChB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;QAClB,CAAC,CAAC,QAAQ,KAAK,KAAK;YAClB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG;YAChB,CAAC,CAAC,QAAQ,KAAK,OAAO;gBACpB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;gBAClB,CAAC,CAAC,QAAQ,KAAK,YAAY;oBACzB,CAAC,CAAC,GAAG,GAAG,CAAC,kBAAkB,IAAI,yCAA4B,IAAI,GAAG,CAAC,MAAM,CAAC,UAAU,EAAE;oBACtF,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC;IAClC,MAAM,GAAG,GAAG,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,EAAE,GAAG,QAAQ,IAAI,CAAC,KAAK,IAAI,SAAS,CAAC,CAAC,OAAO,CAAC,UAAU,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IAC/G,OAAO,IAAI,oBAAS,CAAC,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,GAAG,EAAE,QAAQ,EAAE,GAAG,CAAC,WAAW,EAAE,CAAC,CAAC;AAC1G,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QACrF,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;QACrD,KAAK,OAAO;YACR,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAChE,KAAK,MAAM;YACP,MAAM,IAAI,KAAK,CAAC,qDAAqD,CAAC,CAAC;QAC3E,KAAK,YAAY;YACb,iBAAiB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YAC7B,OAAO,IAAI,0BAAa,CACpB,GAAG,CAAC,KAAK,CAAC,aAAa,EACvB,GAAG,CAAC,kBAAkB,EACtB,GAAG,CAAC,MAAM,CAAC,UAAU,EACrB,GAAG,CAAC,WAAW,CAClB,CAAC;QACN;YACI,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAChB,GAAG,CAAC,KAAK,CAAC,SAAS,EACnB,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,IAAI,OAAO,CAAgB,EAC7C,GAAG,CAAC,QAAQ,IAAI,SAAS,CAC5B,CAAC;IACV,CAAC;AACL,CAAC;AAED,uDAAuD;AACvD,MAAM,aAAa,GAAG,CAAC,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,CAAC,CAAC;AAE5E,sFAAsF;AAC/E,KAAK,UAAU,UAAU,CAAC,QAAwB,EAAE,GAAkB,EAAE,QAAgB;IAC3F,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAA,uBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QAC7F,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAA,qBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QACjE,KAAK,OAAO;YACR,OAAO,IAAA,uBAAe,GAAE,CAAC;QAC7B,KAAK,YAAY,CAAC,CAAC,CAAC;YAChB,6DAA6D;YAC7D,MAAM,KAAK,GAAG,GAAG,CAAC,kBAAkB,IAAI,yCAA4B,CAAC;YACrE,MAAM,MAAM,GAAG,MAAM,IAAA,uCAA0B,EAAC,QAAQ,CAAC,CAAC;YAC1D,OAAO,IAAA,8BAAiB,EAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,KAAK,CAAC,EAAE,MAAM,IAAI,EAAE,EAAE,QAAQ,CAAC,CAAC;QACvF,CAAC;QACD;YACI,OAAO,aAAa,CAAC;IAC7B,CAAC;AACL,CAAC;AAED,kHAAkH;AAC3G,KAAK,UAAU,aAAa,CAAC,QAAwB,EAAE,QAAgB;IAC1E,IAAI,QAAQ,KAAK,YAAY,EAAE,CAAC;QAC5B,OAAO,CAAC,MAAM,IAAA,uCAA0B,EAAC,eAAe,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAC9E,CAAC;IACD,OAAO,QAAQ,KAAK,MAAM,CAAC,CAAC,CAAC,IAAA,qBAAc,EAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;AAC/D,CAAC;AAED,8FAA8F;AACvF,KAAK,UAAU,aAAa,CAAC,QAAwB;IACxD,OAAO,QAAQ,KAAK,YAAY,CAAC,CAAC,CAAC,CAAC,MAAM,IAAA,uCAA0B,EAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;AACxG,CAAC","sourcesContent":["/**\n * Factory that builds the configured STT/TTS engine. Cloud engines (OpenAI/Azure/AWS) use resolved\n * credentials; local engines (Vosk/Piper) download their models on demand into the data dir.\n * Adding a provider = one new engine class + one `case` here.\n */\nimport { OpenAiStt, type SttEngine } from './stt';\nimport { OpenAiTts, type OpenAiVoice, type TtsEngine } from './tts';\nimport { AzureStt, AzureTts, listAzureVoices } from './azure';\nimport { AwsStt, AwsTts, listPollyVoices, type AwsCreds } from './aws';\nimport { VoskStt, listVoskModels } from './vosk';\nimport { PiperTts, listPiperVoices } from './piper';\nimport {\n    OpenRouterStt,\n    OpenRouterTts,\n    listOpenRouterSpeechModels,\n    voicesForLanguage,\n    DEFAULT_OPENROUTER_TTS_MODEL,\n    type AudioDecoder,\n} from './openrouter';\nimport * as path from 'node:path';\n\nimport { CachedTts } from './ttsCache';\nimport { FallbackStt, FallbackTts, type NamedEngine } from './fallback';\nimport { isoToLocale } from './lang';\nimport type { VoiceLogger } from './download';\n\nexport type SpeechProvider = 'openai' | 'openrouter' | 'azure' | 'aws' | 'vosk' | 'piper';\n\n/** All speech credentials, resolved from config (or the central credential store). */\nexport interface VoiceCredentials {\n    openaiKey: string;\n    openrouterKey: string;\n    azureKey: string;\n    azureRegion: string;\n    aws: AwsCreds;\n}\n\n/** Provider-specific voice ids (each provider names voices differently). */\nexport interface VoiceNames {\n    openai: string; // alloy/echo/…\n    azure: string; // e.g. de-DE-KatjaNeural\n    aws: string; // e.g. Vicki\n    piper: string; // e.g. de_DE-thorsten-medium\n    openrouter: string; // depends on the model, e.g. Kore, george\n}\n\n/** Everything the factory needs to build any engine. */\nexport interface EngineContext {\n    creds: VoiceCredentials;\n    voices: VoiceNames;\n    /** Optional second engine, used when the primary one fails ('' = none). */\n    sttFallback?: SpeechProvider | '';\n    ttsFallback?: SpeechProvider | '';\n    /** Writable instance data dir (for the local engines' downloads). */\n    dataDir: string;\n    log: VoiceLogger;\n    /** Local model overrides ('' → default per language). */\n    voskModel: string;\n    /** OpenAI STT model ('' → 'whisper-1'; e.g. 'gpt-4o-transcribe'). */\n    sttModel: string;\n    /** OpenAI TTS model ('' → 'tts-1'; e.g. 'gpt-4o-mini-tts'). */\n    ttsModel: string;\n    /** OpenRouter models ('' → the defaults of openrouter.ts) */\n    openrouterSttModel: string;\n    openrouterTtsModel: string;\n    /** Decoder for compressed speech (OpenRouter models that only return mp3); absent → those models fail */\n    decodeAudio?: AudioDecoder;\n    /** Longest text that is spoken, in characters; longer ones are cut at a sentence end. 0 = no limit, absent = 400. */\n    ttsMaxChars?: number;\n}\n\nfunction requireAzure(creds: VoiceCredentials): void {\n    if (!creds.azureKey || !creds.azureRegion) {\n        throw new Error('Azure speech needs a subscription key and region');\n    }\n}\n\nfunction requireAws(creds: VoiceCredentials): void {\n    if (!creds.aws.accessKeyId || !creds.aws.secretAccessKey || !creds.aws.region) {\n        throw new Error('AWS speech needs an access key id, secret access key and region');\n    }\n}\n\nfunction requireOpenAi(creds: VoiceCredentials): void {\n    if (!creds.openaiKey) {\n        throw new Error('OpenAI speech needs an API key');\n    }\n}\n\nfunction requireOpenRouter(creds: VoiceCredentials): void {\n    if (!creds.openrouterKey) {\n        throw new Error('OpenRouter speech needs an API key');\n    }\n}\n\nexport function createSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    const primary: NamedEngine<SttEngine> = { name: provider, engine: buildSttEngine(provider, ctx) };\n    const fallback = ctx.sttFallback && ctx.sttFallback !== provider ? ctx.sttFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackStt([primary, { name: fallback, engine: buildSttEngine(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        ctx.log.warn(`Speech-to-text fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\nfunction buildSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureStt(ctx.creds.azureKey, ctx.creds.azureRegion);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsStt(ctx.creds.aws);\n        case 'vosk':\n            return new VoskStt(ctx.dataDir, ctx.voskModel, ctx.log);\n        case 'openrouter':\n            requireOpenRouter(ctx.creds);\n            return new OpenRouterStt(ctx.creds.openrouterKey, ctx.openrouterSttModel);\n        case 'piper':\n            throw new Error('Piper is a text-to-speech engine, not speech-to-text');\n        default:\n            requireOpenAi(ctx.creds);\n            return new OpenAiStt(ctx.creds.openaiKey, ctx.sttModel || undefined);\n    }\n}\n\nexport function createTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    // Each engine gets the shared behaviour (disk cache, length limit, SSML) and its own cache directory,\n    // so a fallback engine's voice can never be served later while the primary one is healthy again.\n    const primary: NamedEngine<TtsEngine> = { name: provider, engine: cachedTts(provider, ctx) };\n    const fallback = ctx.ttsFallback && ctx.ttsFallback !== provider ? ctx.ttsFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackTts([primary, { name: fallback, engine: cachedTts(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        // A fallback that cannot even be built (missing key, wrong provider) must not take the main one down.\n        ctx.log.warn(`Text-to-speech fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\n/**\n * One provider binding plus the shared cache/length/SSML layer. One cache directory per provider+voice —\n * a voice change must not serve audio in the old voice.\n */\nfunction cachedTts(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    const voice =\n        provider === 'azure'\n            ? ctx.voices.azure\n            : provider === 'aws'\n              ? ctx.voices.aws\n              : provider === 'piper'\n                ? ctx.voices.piper\n                : provider === 'openrouter'\n                  ? `${ctx.openrouterTtsModel || DEFAULT_OPENROUTER_TTS_MODEL}-${ctx.voices.openrouter}`\n                  : ctx.voices.openai;\n    const dir = path.join(ctx.dataDir, 'ttscache', `${provider}-${(voice || 'default').replace(/[^\\w.-]/g, '_')}`);\n    return new CachedTts(buildTtsEngine(provider, ctx), { dir, log: ctx.log, maxChars: ctx.ttsMaxChars });\n}\n\nfunction buildTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureTts(ctx.creds.azureKey, ctx.creds.azureRegion, ctx.voices.azure);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsTts(ctx.creds.aws, ctx.voices.aws);\n        case 'piper':\n            return new PiperTts(ctx.dataDir, ctx.voices.piper, ctx.log);\n        case 'vosk':\n            throw new Error('Vosk is a speech-to-text engine, not text-to-speech');\n        case 'openrouter':\n            requireOpenRouter(ctx.creds);\n            return new OpenRouterTts(\n                ctx.creds.openrouterKey,\n                ctx.openrouterTtsModel,\n                ctx.voices.openrouter,\n                ctx.decodeAudio,\n            );\n        default:\n            requireOpenAi(ctx.creds);\n            return new OpenAiTts(\n                ctx.creds.openaiKey,\n                (ctx.voices.openai || 'alloy') as OpenAiVoice,\n                ctx.ttsModel || undefined,\n            );\n    }\n}\n\n/** OpenAI TTS voices are a fixed set (no list API). */\nconst OPENAI_VOICES = ['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'];\n\n/** List the available TTS voices for a provider (for the settings voice dropdown). */\nexport async function listVoices(provider: SpeechProvider, ctx: EngineContext, language: string): Promise<string[]> {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return listAzureVoices(ctx.creds.azureKey, ctx.creds.azureRegion, isoToLocale(language));\n        case 'aws':\n            requireAws(ctx.creds);\n            return listPollyVoices(ctx.creds.aws, isoToLocale(language));\n        case 'piper':\n            return listPiperVoices();\n        case 'openrouter': {\n            // public list, no key needed: the voices of the chosen model\n            const model = ctx.openrouterTtsModel || DEFAULT_OPENROUTER_TTS_MODEL;\n            const models = await listOpenRouterSpeechModels('speech');\n            return voicesForLanguage(models.find(m => m.id === model)?.voices || [], language);\n        }\n        default:\n            return OPENAI_VOICES;\n    }\n}\n\n/** Suggested STT model names for the settings dropdown: the local Vosk catalogue, or the models of OpenRouter. */\nexport async function listSttModels(provider: SpeechProvider, language: string): Promise<string[]> {\n    if (provider === 'openrouter') {\n        return (await listOpenRouterSpeechModels('transcription')).map(m => m.id);\n    }\n    return provider === 'vosk' ? listVoskModels(language) : [];\n}\n\n/** TTS model names for the settings dropdown — only OpenRouter has a choice worth listing. */\nexport async function listTtsModels(provider: SpeechProvider): Promise<string[]> {\n    return provider === 'openrouter' ? (await listOpenRouterSpeechModels('speech')).map(m => m.id) : [];\n}\n"]}
+{"version":3,"file":"engines.js","sourceRoot":"","sources":["../../../src/lib/voice/engines.ts"],"names":[],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAsGA,0CAYC;AA0BD,0CAeC;AAuDD,gCAmBC;AAGD,sCAKC;AAGD,sCAEC;AAlPD;;;;GAIG;AACH,+BAAkD;AAClD,+BAAoE;AACpE,mCAA8D;AAC9D,+BAAuE;AACvE,iCAAiD;AACjD,mCAAoD;AACpD,6CAOsB;AACtB,gDAAkC;AAElC,yCAAuC;AACvC,yCAAwE;AACxE,iCAAqC;AAuDrC,SAAS,YAAY,CAAC,KAAuB;IACzC,IAAI,CAAC,KAAK,CAAC,QAAQ,IAAI,CAAC,KAAK,CAAC,WAAW,EAAE,CAAC;QACxC,MAAM,IAAI,KAAK,CAAC,kDAAkD,CAAC,CAAC;IACxE,CAAC;AACL,CAAC;AAED,SAAS,UAAU,CAAC,KAAuB;IACvC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,WAAW,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,eAAe,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,MAAM,EAAE,CAAC;QAC5E,MAAM,IAAI,KAAK,CAAC,iEAAiE,CAAC,CAAC;IACvF,CAAC;AACL,CAAC;AAED,SAAS,aAAa,CAAC,KAAuB;IAC1C,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,CAAC;QACnB,MAAM,IAAI,KAAK,CAAC,gCAAgC,CAAC,CAAC;IACtD,CAAC;AACL,CAAC;AAED,SAAS,iBAAiB,CAAC,KAAuB;IAC9C,IAAI,CAAC,KAAK,CAAC,aAAa,EAAE,CAAC;QACvB,MAAM,IAAI,KAAK,CAAC,oCAAoC,CAAC,CAAC;IAC1D,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAClG,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IAC1G,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,CAAC,CAAC;QACnE,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QACrC,KAAK,MAAM;YACP,OAAO,IAAI,cAAO,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,SAAS,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAC5D,KAAK,YAAY;YACb,iBAAiB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YAC7B,OAAO,IAAI,0BAAa,CAAC,GAAG,CAAC,KAAK,CAAC,aAAa,EAAE,GAAG,CAAC,kBAAkB,CAAC,CAAC;QAC9E,KAAK,OAAO;YACR,MAAM,IAAI,KAAK,CAAC,sDAAsD,CAAC,CAAC;QAC5E;YACI,IAAI,GAAG,CAAC,UAAU,EAAE,IAAI,EAAE,EAAE,CAAC;gBACzB,OAAO,IAAI,eAAS,CAAC,GAAG,CAAC,YAAY,IAAI,EAAE,EAAE,GAAG,CAAC,QAAQ,IAAI,SAAS,EAAE,GAAG,CAAC,UAAU,CAAC,CAAC;YAC5F,CAAC;YACD,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAAC,GAAG,CAAC,KAAK,CAAC,SAAS,EAAE,GAAG,CAAC,QAAQ,IAAI,SAAS,CAAC,CAAC;IAC7E,CAAC;AACL,CAAC;AAED,SAAgB,eAAe,CAAC,QAAwB,EAAE,GAAkB;IACxE,sGAAsG;IACtG,iGAAiG;IACjG,MAAM,OAAO,GAA2B,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC;IAC7F,MAAM,QAAQ,GAAG,GAAG,CAAC,WAAW,IAAI,GAAG,CAAC,WAAW,KAAK,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAC,CAAC,EAAE,CAAC;IACxF,IAAI,CAAC,QAAQ,EAAE,CAAC;QACZ,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;IACD,IAAI,CAAC;QACD,OAAO,IAAI,sBAAW,CAAC,CAAC,OAAO,EAAE,EAAE,IAAI,EAAE,QAAQ,EAAE,MAAM,EAAE,SAAS,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;IACrG,CAAC;IAAC,OAAO,CAAC,EAAE,CAAC;QACT,sGAAsG;QACtG,GAAG,CAAC,GAAG,CAAC,IAAI,CAAC,4BAA4B,QAAQ,kBAAmB,CAAW,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3F,OAAO,OAAO,CAAC,MAAM,CAAC;IAC1B,CAAC;AACL,CAAC;AAED;;;GAGG;AACH,SAAS,SAAS,CAAC,QAAwB,EAAE,GAAkB;IAC3D,MAAM,KAAK,GACP,QAAQ,KAAK,OAAO;QAChB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;QAClB,CAAC,CAAC,QAAQ,KAAK,KAAK;YAClB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,GAAG;YAChB,CAAC,CAAC,QAAQ,KAAK,OAAO;gBACpB,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,KAAK;gBAClB,CAAC,CAAC,QAAQ,KAAK,YAAY;oBACzB,CAAC,CAAC,GAAG,GAAG,CAAC,kBAAkB,IAAI,yCAA4B,IAAI,GAAG,CAAC,MAAM,CAAC,UAAU,EAAE;oBACtF,CAAC,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,CAAC;IAClC,MAAM,GAAG,GAAG,IAAI,CAAC,IAAI,CAAC,GAAG,CAAC,OAAO,EAAE,UAAU,EAAE,GAAG,QAAQ,IAAI,CAAC,KAAK,IAAI,SAAS,CAAC,CAAC,OAAO,CAAC,UAAU,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IAC/G,OAAO,IAAI,oBAAS,CAAC,cAAc,CAAC,QAAQ,EAAE,GAAG,CAAC,EAAE,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,GAAG,EAAE,QAAQ,EAAE,GAAG,CAAC,WAAW,EAAE,CAAC,CAAC;AAC1G,CAAC;AAED,SAAS,cAAc,CAAC,QAAwB,EAAE,GAAkB;IAChE,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,CAAC,CAAC;QACrF,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAI,YAAM,CAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,GAAG,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;QACrD,KAAK,OAAO;YACR,OAAO,IAAI,gBAAQ,CAAC,GAAG,CAAC,OAAO,EAAE,GAAG,CAAC,MAAM,CAAC,KAAK,EAAE,GAAG,CAAC,GAAG,CAAC,CAAC;QAChE,KAAK,MAAM;YACP,MAAM,IAAI,KAAK,CAAC,qDAAqD,CAAC,CAAC;QAC3E,KAAK,YAAY;YACb,iBAAiB,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YAC7B,OAAO,IAAI,0BAAa,CACpB,GAAG,CAAC,KAAK,CAAC,aAAa,EACvB,GAAG,CAAC,kBAAkB,EACtB,GAAG,CAAC,MAAM,CAAC,UAAU,EACrB,GAAG,CAAC,WAAW,CAClB,CAAC;QACN;YACI,aAAa,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACzB,OAAO,IAAI,eAAS,CAChB,GAAG,CAAC,KAAK,CAAC,SAAS,EACnB,CAAC,GAAG,CAAC,MAAM,CAAC,MAAM,IAAI,OAAO,CAAgB,EAC7C,GAAG,CAAC,QAAQ,IAAI,SAAS,CAC5B,CAAC;IACV,CAAC;AACL,CAAC;AAED,uDAAuD;AACvD,MAAM,aAAa,GAAG,CAAC,OAAO,EAAE,MAAM,EAAE,OAAO,EAAE,MAAM,EAAE,MAAM,EAAE,SAAS,CAAC,CAAC;AAE5E,sFAAsF;AAC/E,KAAK,UAAU,UAAU,CAAC,QAAwB,EAAE,GAAkB,EAAE,QAAgB;IAC3F,QAAQ,QAAQ,EAAE,CAAC;QACf,KAAK,OAAO;YACR,YAAY,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACxB,OAAO,IAAA,uBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,QAAQ,EAAE,GAAG,CAAC,KAAK,CAAC,WAAW,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QAC7F,KAAK,KAAK;YACN,UAAU,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YACtB,OAAO,IAAA,qBAAe,EAAC,GAAG,CAAC,KAAK,CAAC,GAAG,EAAE,IAAA,kBAAW,EAAC,QAAQ,CAAC,CAAC,CAAC;QACjE,KAAK,OAAO;YACR,OAAO,IAAA,uBAAe,GAAE,CAAC;QAC7B,KAAK,YAAY,CAAC,CAAC,CAAC;YAChB,6DAA6D;YAC7D,MAAM,KAAK,GAAG,GAAG,CAAC,kBAAkB,IAAI,yCAA4B,CAAC;YACrE,MAAM,MAAM,GAAG,MAAM,IAAA,uCAA0B,EAAC,QAAQ,CAAC,CAAC;YAC1D,OAAO,IAAA,8BAAiB,EAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,KAAK,CAAC,EAAE,MAAM,IAAI,EAAE,EAAE,QAAQ,CAAC,CAAC;QACvF,CAAC;QACD;YACI,OAAO,aAAa,CAAC;IAC7B,CAAC;AACL,CAAC;AAED,kHAAkH;AAC3G,KAAK,UAAU,aAAa,CAAC,QAAwB,EAAE,QAAgB;IAC1E,IAAI,QAAQ,KAAK,YAAY,EAAE,CAAC;QAC5B,OAAO,CAAC,MAAM,IAAA,uCAA0B,EAAC,eAAe,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;IAC9E,CAAC;IACD,OAAO,QAAQ,KAAK,MAAM,CAAC,CAAC,CAAC,IAAA,qBAAc,EAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;AAC/D,CAAC;AAED,8FAA8F;AACvF,KAAK,UAAU,aAAa,CAAC,QAAwB;IACxD,OAAO,QAAQ,KAAK,YAAY,CAAC,CAAC,CAAC,CAAC,MAAM,IAAA,uCAA0B,EAAC,QAAQ,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,CAAC,EAAE,CAAC;AACxG,CAAC","sourcesContent":["/**\n * Factory that builds the configured STT/TTS engine. Cloud engines (OpenAI/Azure/AWS) use resolved\n * credentials; local engines (Vosk/Piper) download their models on demand into the data dir.\n * Adding a provider = one new engine class + one `case` here.\n */\nimport { OpenAiStt, type SttEngine } from './stt';\nimport { OpenAiTts, type OpenAiVoice, type TtsEngine } from './tts';\nimport { AzureStt, AzureTts, listAzureVoices } from './azure';\nimport { AwsStt, AwsTts, listPollyVoices, type AwsCreds } from './aws';\nimport { VoskStt, listVoskModels } from './vosk';\nimport { PiperTts, listPiperVoices } from './piper';\nimport {\n    OpenRouterStt,\n    OpenRouterTts,\n    listOpenRouterSpeechModels,\n    voicesForLanguage,\n    DEFAULT_OPENROUTER_TTS_MODEL,\n    type AudioDecoder,\n} from './openrouter';\nimport * as path from 'node:path';\n\nimport { CachedTts } from './ttsCache';\nimport { FallbackStt, FallbackTts, type NamedEngine } from './fallback';\nimport { isoToLocale } from './lang';\nimport type { VoiceLogger } from './download';\n\nexport type SpeechProvider = 'openai' | 'openrouter' | 'azure' | 'aws' | 'vosk' | 'piper';\n\n/** All speech credentials, resolved from config (or the central credential store). */\nexport interface VoiceCredentials {\n    openaiKey: string;\n    openrouterKey: string;\n    azureKey: string;\n    azureRegion: string;\n    aws: AwsCreds;\n}\n\n/** Provider-specific voice ids (each provider names voices differently). */\nexport interface VoiceNames {\n    openai: string; // alloy/echo/…\n    azure: string; // e.g. de-DE-KatjaNeural\n    aws: string; // e.g. Vicki\n    piper: string; // e.g. de_DE-thorsten-medium\n    openrouter: string; // depends on the model, e.g. Kore, george\n}\n\n/** Everything the factory needs to build any engine. */\nexport interface EngineContext {\n    creds: VoiceCredentials;\n    voices: VoiceNames;\n    /** Optional second engine, used when the primary one fails ('' = none). */\n    sttFallback?: SpeechProvider | '';\n    ttsFallback?: SpeechProvider | '';\n    /** Writable instance data dir (for the local engines' downloads). */\n    dataDir: string;\n    log: VoiceLogger;\n    /** Local model overrides ('' → default per language). */\n    voskModel: string;\n    /** OpenAI STT model ('' → 'whisper-1'; e.g. 'gpt-4o-transcribe'). */\n    sttModel: string;\n    /**\n     * OpenAI-compatible transcription server instead of OpenAI ('' = OpenAI), e.g. a local whisper container.\n     * It gets `sttServerKey`, never the OpenAI key — that one is only for OpenAI.\n     */\n    sttBaseUrl?: string;\n    /** Key for `sttBaseUrl` ('' = the server needs none). */\n    sttServerKey?: string;\n    /** OpenAI TTS model ('' → 'tts-1'; e.g. 'gpt-4o-mini-tts'). */\n    ttsModel: string;\n    /** OpenRouter models ('' → the defaults of openrouter.ts) */\n    openrouterSttModel: string;\n    openrouterTtsModel: string;\n    /** Decoder for compressed speech (OpenRouter models that only return mp3); absent → those models fail */\n    decodeAudio?: AudioDecoder;\n    /** Longest text that is spoken, in characters; longer ones are cut at a sentence end. 0 = no limit, absent = 400. */\n    ttsMaxChars?: number;\n}\n\nfunction requireAzure(creds: VoiceCredentials): void {\n    if (!creds.azureKey || !creds.azureRegion) {\n        throw new Error('Azure speech needs a subscription key and region');\n    }\n}\n\nfunction requireAws(creds: VoiceCredentials): void {\n    if (!creds.aws.accessKeyId || !creds.aws.secretAccessKey || !creds.aws.region) {\n        throw new Error('AWS speech needs an access key id, secret access key and region');\n    }\n}\n\nfunction requireOpenAi(creds: VoiceCredentials): void {\n    if (!creds.openaiKey) {\n        throw new Error('OpenAI speech needs an API key');\n    }\n}\n\nfunction requireOpenRouter(creds: VoiceCredentials): void {\n    if (!creds.openrouterKey) {\n        throw new Error('OpenRouter speech needs an API key');\n    }\n}\n\nexport function createSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    const primary: NamedEngine<SttEngine> = { name: provider, engine: buildSttEngine(provider, ctx) };\n    const fallback = ctx.sttFallback && ctx.sttFallback !== provider ? ctx.sttFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackStt([primary, { name: fallback, engine: buildSttEngine(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        ctx.log.warn(`Speech-to-text fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\nfunction buildSttEngine(provider: SpeechProvider, ctx: EngineContext): SttEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureStt(ctx.creds.azureKey, ctx.creds.azureRegion);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsStt(ctx.creds.aws);\n        case 'vosk':\n            return new VoskStt(ctx.dataDir, ctx.voskModel, ctx.log);\n        case 'openrouter':\n            requireOpenRouter(ctx.creds);\n            return new OpenRouterStt(ctx.creds.openrouterKey, ctx.openrouterSttModel);\n        case 'piper':\n            throw new Error('Piper is a text-to-speech engine, not speech-to-text');\n        default:\n            if (ctx.sttBaseUrl?.trim()) {\n                return new OpenAiStt(ctx.sttServerKey || '', ctx.sttModel || undefined, ctx.sttBaseUrl);\n            }\n            requireOpenAi(ctx.creds);\n            return new OpenAiStt(ctx.creds.openaiKey, ctx.sttModel || undefined);\n    }\n}\n\nexport function createTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    // Each engine gets the shared behaviour (disk cache, length limit, SSML) and its own cache directory,\n    // so a fallback engine's voice can never be served later while the primary one is healthy again.\n    const primary: NamedEngine<TtsEngine> = { name: provider, engine: cachedTts(provider, ctx) };\n    const fallback = ctx.ttsFallback && ctx.ttsFallback !== provider ? ctx.ttsFallback : '';\n    if (!fallback) {\n        return primary.engine;\n    }\n    try {\n        return new FallbackTts([primary, { name: fallback, engine: cachedTts(fallback, ctx) }], ctx.log);\n    } catch (e) {\n        // A fallback that cannot even be built (missing key, wrong provider) must not take the main one down.\n        ctx.log.warn(`Text-to-speech fallback '${fallback}' unavailable: ${(e as Error).message}`);\n        return primary.engine;\n    }\n}\n\n/**\n * One provider binding plus the shared cache/length/SSML layer. One cache directory per provider+voice —\n * a voice change must not serve audio in the old voice.\n */\nfunction cachedTts(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    const voice =\n        provider === 'azure'\n            ? ctx.voices.azure\n            : provider === 'aws'\n              ? ctx.voices.aws\n              : provider === 'piper'\n                ? ctx.voices.piper\n                : provider === 'openrouter'\n                  ? `${ctx.openrouterTtsModel || DEFAULT_OPENROUTER_TTS_MODEL}-${ctx.voices.openrouter}`\n                  : ctx.voices.openai;\n    const dir = path.join(ctx.dataDir, 'ttscache', `${provider}-${(voice || 'default').replace(/[^\\w.-]/g, '_')}`);\n    return new CachedTts(buildTtsEngine(provider, ctx), { dir, log: ctx.log, maxChars: ctx.ttsMaxChars });\n}\n\nfunction buildTtsEngine(provider: SpeechProvider, ctx: EngineContext): TtsEngine {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return new AzureTts(ctx.creds.azureKey, ctx.creds.azureRegion, ctx.voices.azure);\n        case 'aws':\n            requireAws(ctx.creds);\n            return new AwsTts(ctx.creds.aws, ctx.voices.aws);\n        case 'piper':\n            return new PiperTts(ctx.dataDir, ctx.voices.piper, ctx.log);\n        case 'vosk':\n            throw new Error('Vosk is a speech-to-text engine, not text-to-speech');\n        case 'openrouter':\n            requireOpenRouter(ctx.creds);\n            return new OpenRouterTts(\n                ctx.creds.openrouterKey,\n                ctx.openrouterTtsModel,\n                ctx.voices.openrouter,\n                ctx.decodeAudio,\n            );\n        default:\n            requireOpenAi(ctx.creds);\n            return new OpenAiTts(\n                ctx.creds.openaiKey,\n                (ctx.voices.openai || 'alloy') as OpenAiVoice,\n                ctx.ttsModel || undefined,\n            );\n    }\n}\n\n/** OpenAI TTS voices are a fixed set (no list API). */\nconst OPENAI_VOICES = ['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'];\n\n/** List the available TTS voices for a provider (for the settings voice dropdown). */\nexport async function listVoices(provider: SpeechProvider, ctx: EngineContext, language: string): Promise<string[]> {\n    switch (provider) {\n        case 'azure':\n            requireAzure(ctx.creds);\n            return listAzureVoices(ctx.creds.azureKey, ctx.creds.azureRegion, isoToLocale(language));\n        case 'aws':\n            requireAws(ctx.creds);\n            return listPollyVoices(ctx.creds.aws, isoToLocale(language));\n        case 'piper':\n            return listPiperVoices();\n        case 'openrouter': {\n            // public list, no key needed: the voices of the chosen model\n            const model = ctx.openrouterTtsModel || DEFAULT_OPENROUTER_TTS_MODEL;\n            const models = await listOpenRouterSpeechModels('speech');\n            return voicesForLanguage(models.find(m => m.id === model)?.voices || [], language);\n        }\n        default:\n            return OPENAI_VOICES;\n    }\n}\n\n/** Suggested STT model names for the settings dropdown: the local Vosk catalogue, or the models of OpenRouter. */\nexport async function listSttModels(provider: SpeechProvider, language: string): Promise<string[]> {\n    if (provider === 'openrouter') {\n        return (await listOpenRouterSpeechModels('transcription')).map(m => m.id);\n    }\n    return provider === 'vosk' ? listVoskModels(language) : [];\n}\n\n/** TTS model names for the settings dropdown — only OpenRouter has a choice worth listing. */\nexport async function listTtsModels(provider: SpeechProvider): Promise<string[]> {\n    return provider === 'openrouter' ? (await listOpenRouterSpeechModels('speech')).map(m => m.id) : [];\n}\n"]}

package/build/lib/voice/esphome.js

@@ -35,6 +35,7 @@ var __importStar = (this && this.__importStar) || (function () {
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.EsphomeSatellites = exports.TIMER_EVENT = exports.DEFAULT_SPEECH_LEVEL = void 0;
 exports.speechLevelOf = speechLevelOf;
+exports.esphomeAddress = esphomeAddress;
 /**
  * ESPHome voice satellites — the fourth transport, next to the UDP server (`voiceServer.ts`) and the
  * Wyoming endpoint (`wyoming.ts`).
@@ -63,7 +64,7 @@ const esphomeProto_1 = require("./esphomeProto");
 const esphomeEntities_1 = require("./esphomeEntities");
 const vad_1 = require("./vad");
 const speechAudio_1 = require("./speechAudio");
-const stt_1 = require("./stt");
+const replyFormat_1 = require("./replyFormat");
 const captures_1 = require("./captures");
 /** `VoiceAssistantEvent` (api.proto) — the subset we emit. */
 const EVENT = {
@@ -154,6 +155,16 @@ const STREAM_START_MS = 5000;
 const STALL_CHECK_MS = 250;
 /** Silence threshold as a share of the speech threshold — keeps the default pair 250/150. */
 const SILENCE_SHARE = 0.6;
+/**
+ * The firmware answers an ERROR event with a stop request of its own (`signal_stop_`). One arriving this soon after
+ * our error belongs to it and needs no reply — the run is already over.
+ */
+const ERROR_ECHO_MS = 2000;
+/**
+ * What ends a run that produced no request. Home Assistant sends the same code when the speech recognition finds no
+ * text, and the Voice PE firmware deliberately plays no error sound for it.
+ */
+const NO_TEXT_ERROR = { code: 'stt-no-text-recognized', message: 'No text recognized' };
 /**
  * A dropped utterance whose loudest frame reached this share of the speech threshold was probably a voice
  * that was too quiet, not an empty room — worth telling the user where to lower the threshold.
@@ -164,6 +175,12 @@ function speechLevelOf(device) {
     const level = Math.round(Number(device.speechLevel));
     return Number.isFinite(level) && level > 0 ? level : 0;
 }
+/** How a configured device is identified: `ip`, or `ip:port` when it does not use the default port. */
+function esphomeAddress(device) {
+    const ip = (device.ip || '').trim().toLowerCase();
+    const port = Number(device.port) || DEFAULT_PORT;
+    return port === DEFAULT_PORT ? ip : `${ip}:${port}`;
+}
 /** `VoiceAssistantTimerEvent` in api.proto. */
 exports.TIMER_EVENT = { started: 0, updated: 1, cancelled: 2, finished: 3 };
 /**
@@ -293,13 +310,29 @@ class EsphomeConnection {
     password;
     /** Configured speech threshold, 0 = {@link DEFAULT_SPEECH_LEVEL}. */
     speechLevel;
+    /** `ip`, or `ip:port` for another port than the default — how the adapter recognises the table row. */
+    address;
+    /** The device's friendly name, for entities without a name of their own. */
+    friendlyName = '';
+    /**
+     * Number of the current pipeline run. A run whose number is no longer current was ended — the device asked to
+     * stop, started over, or went away — and whatever STT, answer or TTS is still in flight for it stays silent.
+     */
+    run = 0;
+    /** From RUN_START until we sent RUN_END: the device waits for that RUN_END. */
+    runOpen = false;
+    /** When we last sent an ERROR event (see {@link ERROR_ECHO_MS}). */
+    errorSentAt = 0;
+    /** The reply format last reported in the log, to log it again only when it changes. */
+    loggedFormat = '';
     constructor(device, opts) {
         this.opts = opts;
         this.host = device.ip.trim();
-        this.port = device.port || DEFAULT_PORT;
+        this.port = Number(device.port) || DEFAULT_PORT;
         this.room = (device.room || '').trim();
         this.password = device.password || '';
         this.speechLevel = speechLevelOf(device);
+        this.address = esphomeAddress(device);
         this.deviceName = this.host;
     }
     // ── lifecycle ───────────────────────────────────────────────────────────────────────────────────
@@ -355,6 +388,9 @@ class EsphomeConnection {
     onDisconnected() {
         this.clearTimers();
         this.abortCapture();
+        // A reply still being worked out must not land on the next connection, in the middle of another run.
+        this.run++;
+        this.runOpen = false;
         if (this.online) {
             this.opts.log.info(`ESPHome satellite offline: ${this.deviceName}`);
         }
@@ -388,11 +424,11 @@ class EsphomeConnection {
     setOffline() {
         if (this.online) {
             this.online = false;
-            this.opts.onStatus?.(this.deviceName, this.room, 'offline');
+            this.opts.onStatus?.(this.deviceName, this.room, 'offline', this.address);
         }
     }
     status(state) {
-        this.opts.onStatus?.(this.deviceName, this.room, state);
+        this.opts.onStatus?.(this.deviceName, this.room, state, this.address);
     }
     // ── protocol plumbing ───────────────────────────────────────────────────────────────────────────
     send(message) {
@@ -479,6 +515,7 @@ class EsphomeConnection {
             case 'DeviceInfoResponse': {
                 const info = message;
                 this.deviceName = info.getName() || this.host;
+                this.friendlyName = info.getFriendlyName?.() || info.getName() || '';
                 this.online = true;
                 this.opts.log.info(`ESPHome satellite online: ${this.deviceName} (${info.getModel()}) @ ${this.host}${this.room ? `, room ${this.room}` : ''}`);
                 this.status('idle');
@@ -556,7 +593,7 @@ class EsphomeConnection {
             }
             default:
                 if ((0, esphomeEntities_1.isListEntityMessage)(name)) {
-                    this.entities.addFromAnnouncement(name, message);
+                    this.entities.addFromAnnouncement(name, message, this.friendlyName);
                 }
                 else if ((0, esphomeEntities_1.isEntityStateMessage)(name)) {
                     const entity = this.entities.applyState(name, message);
@@ -600,13 +637,19 @@ class EsphomeConnection {
     // ── voice pipeline ──────────────────────────────────────────────────────────────────────────────
     onPipelineRequest(request) {
         if (!request.getStart()) {
-            this.abortCapture();
+            this.onStopRequest();
             return;
         }
         if (this.capture) {
             this.opts.log.debug(`ESPHome ${this.deviceName}: pipeline already running — ignoring wake`);
             return;
         }
+        if (this.runOpen) {
+            // The device started over while the last request was still being answered: that answer is not
+            // wanted any more, and its events would land in the middle of the new run.
+            this.opts.log.debug(`ESPHome ${this.deviceName}: new wake while answering — dropping the old reply`);
+        }
+        this.run++;
         const phrase = request.getWakeWordPhrase();
         this.opts.log.debug(`ESPHome ${this.deviceName}: wake "${phrase || '(follow-up)'}"`);
         const response = new esphomeProto_1.pb.VoiceAssistantResponse();
@@ -653,6 +696,29 @@ class EsphomeConnection {
         this.event(EVENT.STT_START);
         this.event(EVENT.STT_VAD_START);
     }
+    /**
+     * The device asks to end the conversation (`VoiceAssistantRequest{start:false}`): a button, or the wake word
+     * said again while a request is running — that is how ESPHome's voice_assistant stops. While it waits for an
+     * answer, the firmware leaves that state only on our RUN_END; without one it stays "in the conversation", and
+     * every further wake word just sends another stop (field report 2026-11-03).
+     */
+    onStopRequest() {
+        const running = !!this.capture || this.runOpen;
+        this.abortCapture();
+        if (running) {
+            this.run++; // whatever is still being answered stays silent
+            this.opts.log.debug(`ESPHome ${this.deviceName}: the device ended the conversation`);
+            this.event(EVENT.RUN_END);
+            this.status('idle');
+            return;
+        }
+        if (Date.now() - this.errorSentAt < ERROR_ECHO_MS) {
+            return; // the device's own reaction to our ERROR event
+        }
+        // Nothing is open on our side, but the device still thinks so (it missed a RUN_END): end it there too.
+        this.opts.log.debug(`ESPHome ${this.deviceName}: stop outside a conversation — ending it on the device`);
+        this.event(EVENT.RUN_END);
+    }
     onAudio(audio) {
         const capture = this.capture;
         if (!capture || capture.closed) {
@@ -718,6 +784,15 @@ class EsphomeConnection {
         this.capture = null;
         clearTimeout(capture.timer);
         clearInterval(capture.stallCheck);
+        const run = this.run;
+        /** False once the run was ended elsewhere — then nothing more may be sent for it. */
+        const current = () => {
+            if (run === this.run) {
+                return true;
+            }
+            this.opts.log.debug(`ESPHome ${this.deviceName}: conversation ended meanwhile — dropping the reply`);
+            return false;
+        };
         // Send this first: the device streams until it sees it.
         this.event(EVENT.STT_VAD_END);
         const pcm = Buffer.concat(capture.chunks);
@@ -747,13 +822,15 @@ class EsphomeConnection {
                 this.opts.log.debug(`ESPHome ${this.deviceName}: nothing but silence${end === 'no-speech' ? ' (nobody started speaking)' : ''} — dropping the utterance`);
             }
             this.discard(capture, pcm, end === 'no-speech' ? 'no-speech' : 'silence', vad);
-            this.event(EVENT.RUN_END);
-            this.status('idle');
+            this.endWithoutRequest();
             return;
         }
         this.status('processing');
         try {
             const hints = this.opts.getHints ? await this.opts.getHints() : undefined;
+            if (!current()) {
+                return;
+            }
             const speech = (0, speechAudio_1.prepareForStt)(pcm, MIC_SAMPLE_RATE, { startMs: vad.speechStartMs, endLevel: vad.endLevel });
             vad.stt = { fromMs: speech.fromMs, toMs: speech.toMs, gain: speech.gain };
             this.opts.log.debug(`ESPHome ${this.deviceName}: speech recognition gets ${(speech.fromMs / 1000).toFixed(2)}–` +
@@ -761,16 +838,19 @@ class EsphomeConnection {
             const sttStart = Date.now();
             const text = (await this.opts.stt.transcribe(speech.pcm, MIC_SAMPLE_RATE, this.opts.language, hints)).trim();
             const sttMs = Date.now() - sttStart;
-            this.event(EVENT.STT_END, { text });
+            if (!current()) {
+                return;
+            }
             if (!text) {
                 // Speech was heard, so the user waits for an answer: say why none comes. Typically a word in another
                 // language than the STT listens in (a German "Ja" to a Russian STT comes back empty).
                 this.opts.log.info(`Voice Q (${this.deviceName}): (empty) — speech recognition returned no text for ${seconds.toFixed(1)} s of audio`);
                 this.discard(capture, pcm, 'empty-stt', vad);
-                this.event(EVENT.RUN_END);
-                this.status('idle');
+                // No STT_END without text — like Home Assistant, which reports this case as an error only.
+                this.endWithoutRequest();
                 return;
             }
+            this.event(EVENT.STT_END, { text });
             this.event(EVENT.INTENT_START);
             const answer = (await this.opts.answer(text, {
                 device: this.deviceName,
@@ -781,6 +861,9 @@ class EsphomeConnection {
                 // Numbers only, no audio: shows where a request was cut off without keeping recordings.
                 vad: { ...vad, levels: (0, captures_1.levelProfile)(pcm, MIC_SAMPLE_RATE) },
             })).trim();
+            if (!current()) {
+                return;
+            }
             const followUp = !!answer && this.opts.expectsFollowUp?.(answer, this.deviceName) === true;
             this.event(EVENT.INTENT_END, followUp ? { continue_conversation: '1' } : undefined);
             if (!answer) {
@@ -790,11 +873,17 @@ class EsphomeConnection {
             }
             const ttsStart = Date.now();
             const { pcm: ttsPcm, sampleRate } = await this.opts.tts.synthesize(answer, this.opts.language);
+            if (!current()) {
+                return;
+            }
             this.opts.onSpoken?.(this.deviceName, {
                 ms: Date.now() - ttsStart,
                 audioMs: (ttsPcm.length / 2 / sampleRate) * 1000,
             });
-            const url = this.publish(ttsPcm, sampleRate);
+            const url = await this.publish(ttsPcm, sampleRate);
+            if (!current()) {
+                return;
+            }
             this.status('speaking');
             this.event(EVENT.TTS_START, { text: answer });
             this.event(EVENT.TTS_END, { url });
@@ -804,12 +893,26 @@ class EsphomeConnection {
         catch (e) {
             const reason = e.message;
             this.opts.log.error(`ESPHome ${this.deviceName}: voice pipeline failed — ${reason}`);
+            if (run !== this.run) {
+                return; // the conversation is over already — nothing to tell the device
+            }
             // Without this the device stays ducked and its LED keeps spinning.
             this.event(EVENT.ERROR, { code: 'pipeline-failed', message: reason });
             this.event(EVENT.RUN_END);
             this.status('idle');
         }
     }
+    /**
+     * End a run that produced no request. A RUN_END right after STT_VAD_END gets lost: the firmware is still stopping
+     * its microphone then, ignores RUN_END in that state, and afterwards waits for an answer that never comes — the
+     * satellite stays in the conversation (field report 2026-11-03). An ERROR event puts it back to idle in any state,
+     * so it goes first, with the code Home Assistant uses for this case.
+     */
+    endWithoutRequest() {
+        this.event(EVENT.ERROR, NO_TEXT_ERROR);
+        this.event(EVENT.RUN_END);
+        this.status('idle');
+    }
     /**
      * Select which wake words listen on this device. Unknown ids are dropped and the list is trimmed
      * to what the firmware allows, because the device answers a rejected list with silence rather than
@@ -886,12 +989,12 @@ class EsphomeConnection {
      * Speak a ready PCM buffer on this device (timers, alarms, `tts.text`, jingles). With
      * `startConversation` the microphone opens once the clip has played, to catch the answer to a question.
      */
-    announce(pcm, sampleRate, startConversation = false) {
+    async announce(pcm, sampleRate, startConversation = false) {
         if (!this.online || !pcm.length) {
-            return Promise.resolve();
+            return;
         }
         const request = new esphomeProto_1.pb.VoiceAssistantAnnounceRequest();
-        request.setMediaId(this.publish(pcm, sampleRate));
+        request.setMediaId(await this.publish(pcm, sampleRate));
         request.setText('');
         request.setPreannounceMediaId('');
         // Needs feature flag 32 (START_CONVERSATION); on a device without it the clip still plays and
@@ -900,16 +1003,45 @@ class EsphomeConnection {
         this.status('speaking');
         this.send(request);
         this.armPlaybackTimeout();
-        return Promise.resolve();
+    }
+    /** The format replies should have on this device, from what its media player announced. */
+    replyFormat() {
+        const formats = this.entities
+            .list()
+            .filter(e => e.kind === 'mediaPlayer')
+            .flatMap(e => e.formats ?? []);
+        return (0, replyFormat_1.chooseReplyFormat)(formats, !!this.opts.encodeAudio);
     }
     /**
-     * Publish PCM as a WAV on the media server and build the URL the device should fetch.
+     * Publish PCM on the media server, in the format the device plays, and build the URL it should fetch.
      *
      * The host matters: on a multi-homed ioBroker machine the device can usually only reach one of the
      * addresses, and the local end of *this* socket is by definition one it can talk to.
      */
-    publish(pcm, sampleRate) {
-        return this.mediaUrl(this.opts.media.publish((0, stt_1.pcmToWav)(pcm, sampleRate), 'audio/wav'));
+    async publish(pcm, sampleRate) {
+        let format = this.replyFormat();
+        let encoded;
+        try {
+            encoded = await (0, replyFormat_1.encodeReply)(pcm, sampleRate, format, this.opts.encodeAudio);
+        }
+        catch (e) {
+            // MP3/Opus need ffmpeg. Without it a native format is the only chance, whether the device lists it or not.
+            const fallback = { ...format, codec: 'flac', source: `${format.source}, FLAC as fallback` };
+            this.opts.log.warn(`ESPHome ${this.deviceName}: cannot write the reply as ${format.codec.toUpperCase()} ` +
+                `(${e.message}) — sending FLAC, which the device may not play. Install ffmpeg.`);
+            format = fallback;
+            encoded = await (0, replyFormat_1.encodeReply)(pcm, sampleRate, format);
+        }
+        const key = `${format.codec}/${format.sampleRate}/${format.channels}`;
+        if (key !== this.loggedFormat) {
+            this.loggedFormat = key;
+            const what = !format.source
+                ? 'WAV (the device lists no formats)'
+                : `${format.codec.toUpperCase()}${format.sampleRate ? ` ${format.sampleRate} Hz` : ''}` +
+                    `${format.channels === 2 ? ' stereo' : ''} (device: ${format.source})`;
+            this.opts.log.info(`ESPHome ${this.deviceName}: spoken replies as ${what}`);
+        }
+        return this.mediaUrl(this.opts.media.publish(encoded.body, encoded.contentType));
     }
     /** Absolute URL of a path on the media server, with a host this device can reach. */
     mediaUrl(path) {
@@ -931,6 +1063,15 @@ class EsphomeConnection {
         this.playbackTimer.unref();
     }
     event(type, data) {
+        if (type === EVENT.RUN_START) {
+            this.runOpen = true;
+        }
+        else if (type === EVENT.RUN_END) {
+            this.runOpen = false;
+        }
+        else if (type === EVENT.ERROR) {
+            this.errorSentAt = Date.now();
+        }
         const message = new esphomeProto_1.pb.VoiceAssistantEventResponse();
         message.setEventType(type);
         const items = [];