@getpaseo/server 0.1.16 → 0.1.17

package/dist/src/server/speech/providers/openai/realtime-transcription-session.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"realtime-transcription-session.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/realtime-transcription-session.ts"],"names":[],"mappings":"AACA,OAAO,SAAS,MAAM,IAAI,CAAC;AAC3B,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AA6D3C,MAAM,OAAO,kCACX,SAAQ,YAAY;IAgBpB,YAAY,MAOX;QACC,KAAK,EAAE,CAAC;QArBM,uBAAkB,GAAG,KAAK,CAAC;QAQnC,OAAE,GAAqB,IAAI,CAAC;QAC5B,UAAK,GAAyB,IAAI,CAAC;QACnC,YAAO,GAAG,KAAK,CAAC;QAChB,oBAAe,GAAG,IAAI,GAAG,EAAkB,CAAC;QAWlD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,QAAQ,EAAE,SAAS,EAAE,wBAAwB,EAAE,CAAC,CAAC;QAC/F,IAAI,CAAC,kBAAkB,GAAG,MAAM,CAAC,kBAAkB,CAAC;QACpD,IAAI,CAAC,QAAQ,GAAG,MAAM,CAAC,QAAQ,CAAC;QAChC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC;QAC5B,IAAI,CAAC,aAAa,GAAG,MAAM,CAAC,aAAa,IAAI,IAAI,CAAC;IACpD,CAAC;IAEM,KAAK,CAAC,OAAO;QAClB,IAAI,IAAI,CAAC,KAAK,EAAE,CAAC;YACf,OAAO,IAAI,CAAC,KAAK,CAAC;QACpB,CAAC;QAED,IAAI,CAAC,OAAO,GAAG,KAAK,CAAC;QACrB,IAAI,CAAC,KAAK,GAAG,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YACjD,MAAM,GAAG,GAAG,uDAAuD,CAAC;YACpE,MAAM,EAAE,GAAG,IAAI,SAAS,CAAC,GAAG,EAAE;gBAC5B,OAAO,EAAE;oBACP,aAAa,EAAE,UAAU,IAAI,CAAC,MAAM,EAAE;iBACvC;aACF,CAAC,CAAC;YACH,IAAI,CAAC,EAAE,GAAG,EAAE,CAAC;YAEb,IAAI,QAAQ,GAAG,KAAK,CAAC;YAErB,MAAM,IAAI,GAAG,CAAC,KAAY,EAAE,EAAE;gBAC5B,IAAI,QAAQ,EAAE,CAAC;oBACb,IAAI,CAAC,IAAI,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC;oBAC1B,OAAO;gBACT,CAAC;gBACD,QAAQ,GAAG,IAAI,CAAC;gBAChB,MAAM,CAAC,KAAK,CAAC,CAAC;YAChB,CAAC,CAAC;YAEF,EAAE,CAAC,EAAE,CAAC,MAAM,EAAE,GAAG,EAAE;gBACjB,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,mDAAmD,CAAC,CAAC;gBACvE,MAAM,MAAM,GAAsB;oBAChC,IAAI,EAAE,gBAAgB;oBACtB,OAAO,EAAE;wBACP,IAAI,EAAE,eAAe;wBACrB,KAAK,EAAE;4BACL,KAAK,EAAE;gCACL,MAAM,EAAE,EAAE,IAAI,EAAE,WAAW,EAAE,IAAI,EAAE,KAAK,EAAE;gCAC1C,aAAa,EAAE;oCACb,KAAK,EAAE,IAAI,CAAC,kBAAkB;oCAC9B,GAAG,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,IAAI,CAAC,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oCACrD,GAAG,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;iCAChD;gCACD,cAAc,EAAE,IAAI,CAAC,aAAa;6BACnC;yBACF;qBACF;iBACF,CAAC;gBACF,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC;YAClC,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,IAAI,EAAE,EAAE;gBACxB,MAAM,IAAI,GAAG,OAAO,IAAI,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,OAAO,CAAC,CAAC;gBACtE,IAAI,MAAe,CAAC;gBACpB,IAAI,CAAC;oBACH,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gBAC5B,CAAC;gBAAC,MAAM,CAAC;oBACP,OAAO;gBACT,CAAC;gBAED,MAAM,KAAK,GAAG,MAA2B,CAAC;gBAC1C,IACE,KAAK,CAAC,IAAI,KAAK,iBAAiB;oBAChC,KAAK,CAAC,IAAI,KAAK,iBAAiB,EAChC,CAAC;oBACD,IAAI,CAAC,QAAQ,EAAE,CAAC;wBACd,QAAQ,GAAG,IAAI,CAAC;wBAChB,OAAO,EAAE,CAAC;oBACZ,CAAC;oBACD,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,8BAA8B,EAAE,CAAC;oBAClD,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE;wBACrB,SAAS,EAAE,KAAK,CAAC,OAAO;wBACxB,iBAAiB,EAAE,KAAK,CAAC,gBAAgB;qBAC1C,CAAC,CAAC;oBACH,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,mCAAmC,EAAE,CAAC;oBACvD,IAAI,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;oBAC5B,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,mCAAmC,EAAE,CAAC;oBACvD,IAAI,CAAC,IAAI,CAAC,gBAAgB,CAAC,CAAC;oBAC5B,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,mDAAmD,EAAE,CAAC;oBACvE,MAAM,YAAY,GAAG,IAAI,CAAC,kBAAkB,KAAK,WAAW,CAAC;oBAC7D,MAAM,IAAI,GAAG,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;oBAC3D,MAAM,IAAI,GAAG,YAAY,CAAC,CAAC,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,GAAG,KAAK,CAAC,KAAK,CAAC;oBAC7D,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;oBAC9C,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,KAAK,CAAC,OAAO,EAAE,UAAU,EAAE,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC,CAAC;oBACxF,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,uDAAuD,EAAE,CAAC;oBAC3E,IAAI,CAAC,eAAe,CAAC,GAAG,CAAC,KAAK,CAAC,OAAO,EAAE,KAAK,CAAC,UAAU,CAAC,CAAC;oBAC1D,IAAI,CAAC,IAAI,CAAC,YAAY,EAAE,EAAE,SAAS,EAAE,KAAK,CAAC,OAAO,EAAE,UAAU,EAAE,KAAK,CAAC,UAAU,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;oBACnG,OAAO;gBACT,CAAC;gBAED,IAAI,KAAK,CAAC,IAAI,KAAK,OAAO,EAAE,CAAC;oBAC3B,MAAM,OAAO,GAAG,KAAK,CAAC,KAAK,EAAE,OAAO,IAAI,uBAAuB,CAAC;oBAChE,IAAI,CAAC,IAAI,KAAK,CAAC,OAAO,CAAC,CAAC,CAAC;gBAC3B,CAAC;YACH,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,CAAC,GAAG,EAAE,EAAE;gBACrB,IAAI,CAAC,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC;YAC5D,CAAC,CAAC,CAAC;YAEH,EAAE,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;gBAClB,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,kCAAkC,CAAC,CAAC;gBACtD,IAAI,IAAI,CAAC,OAAO,EAAE,CAAC;oBACjB,OAAO;gBACT,CAAC;gBACD,IAAI,CAAC,QAAQ,EAAE,CAAC;oBACd,IAAI,CAAC,IAAI,KAAK,CAAC,+CAA+C,CAAC,CAAC,CAAC;oBACjE,OAAO;gBACT,CAAC;gBACD,IAAI,CAAC,IAAI,KAAK,CAAC,kCAAkC,CAAC,CAAC,CAAC;YACtD,CAAC,CAAC,CAAC;QACL,CAAC,CAAC,CAAC;QAEH,OAAO,IAAI,CAAC,KAAK,CAAC;IACpB,CAAC;IAEM,WAAW,CAAC,OAAe;QAChC,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YACtD,MAAM,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC;QAC7D,CAAC;QACD,MAAM,WAAW,GAAG,OAAO,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAC;QAC/C,MAAM,KAAK,GAAsB,EAAE,IAAI,EAAE,2BAA2B,EAAE,KAAK,EAAE,WAAW,EAAE,CAAC;QAC3F,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;IACtC,CAAC;IAEM,MAAM;QACX,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YACtD,MAAM,IAAI,KAAK,CAAC,yCAAyC,CAAC,CAAC;QAC7D,CAAC;QACD,MAAM,KAAK,GAAsB,EAAE,IAAI,EAAE,2BAA2B,EAAE,CAAC;QACvE,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;IACtC,CAAC;IAEM,KAAK;QACV,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,IAAI,CAAC,EAAE,CAAC,UAAU,KAAK,SAAS,CAAC,IAAI,EAAE,CAAC;YACtD,OAAO;QACT,CAAC;QACD,MAAM,KAAK,GAAsB,EAAE,IAAI,EAAE,0BAA0B,EAAE,CAAC;QACtE,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,KAAK,CAAC,CAAC,CAAC;IACtC,CAAC;IAEM,KAAK;QACV,IAAI,CAAC;YACH,IAAI,CAAC,OAAO,GAAG,IAAI,CAAC;YACpB,IAAI,CAAC,EAAE,EAAE,KAAK,EAAE,CAAC;QACnB,CAAC;QAAC,MAAM,CAAC;YACP,QAAQ;QACV,CAAC;gBAAS,CAAC;YACT,IAAI,CAAC,EAAE,GAAG,IAAI,CAAC;YACf,IAAI,CAAC,KAAK,GAAG,IAAI,CAAC;QACpB,CAAC;IACH,CAAC;CACF"}

package/dist/src/server/speech/providers/openai/runtime.js

@@ -0,0 +1,114 @@
+import { DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL, DEFAULT_OPENAI_TTS_MODEL, } from "./config.js";
+import { OpenAIRealtimeTranscriptionSession } from "./realtime-transcription-session.js";
+import { OpenAISTT } from "./stt.js";
+import { OpenAITTS } from "./tts.js";
+function resolveOpenAiCredentials(openaiConfig) {
+    const openaiApiKey = openaiConfig?.apiKey;
+    return {
+        openaiSttApiKey: openaiConfig?.stt?.apiKey ?? openaiApiKey,
+        openaiTtsApiKey: openaiConfig?.tts?.apiKey ?? openaiApiKey,
+        openaiDictationApiKey: openaiApiKey,
+    };
+}
+export function getOpenAiSpeechAvailability(openaiConfig) {
+    const credentials = resolveOpenAiCredentials(openaiConfig);
+    return {
+        stt: Boolean(credentials.openaiSttApiKey),
+        tts: Boolean(credentials.openaiTtsApiKey),
+        dictationStt: Boolean(credentials.openaiDictationApiKey),
+    };
+}
+export function validateOpenAiCredentialRequirements(params) {
+    const { providers, logger, openaiConfig } = params;
+    const openAiCredentials = resolveOpenAiCredentials(openaiConfig);
+    const missingOpenAiCredentialsFor = [];
+    if (providers.voiceStt.enabled !== false &&
+        providers.voiceStt.provider === "openai" &&
+        !openAiCredentials.openaiSttApiKey) {
+        missingOpenAiCredentialsFor.push("voice.stt");
+    }
+    if (providers.voiceTts.enabled !== false &&
+        providers.voiceTts.provider === "openai" &&
+        !openAiCredentials.openaiTtsApiKey) {
+        missingOpenAiCredentialsFor.push("voice.tts");
+    }
+    if (providers.dictationStt.enabled !== false &&
+        providers.dictationStt.provider === "openai" &&
+        !openAiCredentials.openaiDictationApiKey) {
+        missingOpenAiCredentialsFor.push("dictation.stt");
+    }
+    if (missingOpenAiCredentialsFor.length > 0) {
+        logger.error({
+            requestedProviders: {
+                dictationStt: providers.dictationStt.provider,
+                voiceStt: providers.voiceStt.provider,
+                voiceTts: providers.voiceTts.provider,
+            },
+            missingOpenAiCredentialsFor,
+        }, "Invalid speech configuration: OpenAI provider selected but credentials are missing");
+        throw new Error(`Missing OpenAI credentials for configured speech features: ${missingOpenAiCredentialsFor.join(", ")}`);
+    }
+}
+export function initializeOpenAiSpeechServices(params) {
+    const { providers, openaiConfig, existing, logger } = params;
+    const openAiCredentials = resolveOpenAiCredentials(openaiConfig);
+    let sttService = existing.sttService;
+    let ttsService = existing.ttsService;
+    let dictationSttService = existing.dictationSttService;
+    const needsOpenAiStt = !sttService &&
+        providers.voiceStt.enabled !== false &&
+        providers.voiceStt.provider === "openai";
+    const needsOpenAiTts = !ttsService &&
+        providers.voiceTts.enabled !== false &&
+        providers.voiceTts.provider === "openai";
+    const needsOpenAiDictation = !dictationSttService &&
+        providers.dictationStt.enabled !== false &&
+        providers.dictationStt.provider === "openai";
+    if ((needsOpenAiStt || needsOpenAiTts || needsOpenAiDictation) &&
+        (openAiCredentials.openaiSttApiKey ||
+            openAiCredentials.openaiTtsApiKey ||
+            openAiCredentials.openaiDictationApiKey)) {
+        logger.info("OpenAI speech provider initialized");
+        if (needsOpenAiStt && openAiCredentials.openaiSttApiKey) {
+            const { apiKey: _sttApiKey, ...sttConfig } = openaiConfig?.stt ?? {};
+            sttService = new OpenAISTT({
+                apiKey: openAiCredentials.openaiSttApiKey,
+                ...sttConfig,
+            }, logger);
+        }
+        if (needsOpenAiTts && openAiCredentials.openaiTtsApiKey) {
+            const { apiKey: _ttsApiKey, ...ttsConfig } = openaiConfig?.tts ?? {};
+            ttsService = new OpenAITTS({
+                apiKey: openAiCredentials.openaiTtsApiKey,
+                voice: "alloy",
+                model: DEFAULT_OPENAI_TTS_MODEL,
+                responseFormat: "pcm",
+                ...ttsConfig,
+            }, logger);
+        }
+        const dictationApiKey = openAiCredentials.openaiDictationApiKey;
+        if (needsOpenAiDictation && dictationApiKey) {
+            dictationSttService = {
+                id: "openai",
+                createSession: ({ logger: sessionLogger, language, prompt }) => new OpenAIRealtimeTranscriptionSession({
+                    apiKey: dictationApiKey,
+                    logger: sessionLogger,
+                    transcriptionModel: openaiConfig?.realtimeTranscriptionModel ??
+                        DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL,
+                    ...(language ? { language } : {}),
+                    ...(prompt ? { prompt } : {}),
+                    turnDetection: null,
+                }),
+            };
+        }
+    }
+    else if (needsOpenAiStt || needsOpenAiTts || needsOpenAiDictation) {
+        logger.warn("OpenAI speech providers are configured but credentials are missing");
+    }
+    return {
+        sttService,
+        ttsService,
+        dictationSttService,
+    };
+}
+//# sourceMappingURL=runtime.js.map

package/dist/src/server/speech/providers/openai/runtime.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"runtime.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/runtime.ts"],"names":[],"mappings":"AAIA,OAAO,EACL,2CAA2C,EAC3C,wBAAwB,GAEzB,MAAM,aAAa,CAAC;AACrB,OAAO,EAAE,kCAAkC,EAAE,MAAM,qCAAqC,CAAC;AACzF,OAAO,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AACrC,OAAO,EAAE,SAAS,EAAE,MAAM,UAAU,CAAC;AAoBrC,SAAS,wBAAwB,CAC/B,YAAoD;IAEpD,MAAM,YAAY,GAAG,YAAY,EAAE,MAAM,CAAC;IAC1C,OAAO;QACL,eAAe,EAAE,YAAY,EAAE,GAAG,EAAE,MAAM,IAAI,YAAY;QAC1D,eAAe,EAAE,YAAY,EAAE,GAAG,EAAE,MAAM,IAAI,YAAY;QAC1D,qBAAqB,EAAE,YAAY;KACpC,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,2BAA2B,CACzC,YAAoD;IAEpD,MAAM,WAAW,GAAG,wBAAwB,CAAC,YAAY,CAAC,CAAC;IAC3D,OAAO;QACL,GAAG,EAAE,OAAO,CAAC,WAAW,CAAC,eAAe,CAAC;QACzC,GAAG,EAAE,OAAO,CAAC,WAAW,CAAC,eAAe,CAAC;QACzC,YAAY,EAAE,OAAO,CAAC,WAAW,CAAC,qBAAqB,CAAC;KACzD,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,oCAAoC,CAAC,MAIpD;IACC,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,CAAC;IACnD,MAAM,iBAAiB,GAAG,wBAAwB,CAAC,YAAY,CAAC,CAAC;IAEjE,MAAM,2BAA2B,GAAa,EAAE,CAAC;IACjD,IACE,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;QACpC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;QACxC,CAAC,iBAAiB,CAAC,eAAe,EAClC,CAAC;QACD,2BAA2B,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IAChD,CAAC;IACD,IACE,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;QACpC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ;QACxC,CAAC,iBAAiB,CAAC,eAAe,EAClC,CAAC;QACD,2BAA2B,CAAC,IAAI,CAAC,WAAW,CAAC,CAAC;IAChD,CAAC;IACD,IACE,SAAS,CAAC,YAAY,CAAC,OAAO,KAAK,KAAK;QACxC,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ;QAC5C,CAAC,iBAAiB,CAAC,qBAAqB,EACxC,CAAC;QACD,2BAA2B,CAAC,IAAI,CAAC,eAAe,CAAC,CAAC;IACpD,CAAC;IAED,IAAI,2BAA2B,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QAC3C,MAAM,CAAC,KAAK,CACV;YACE,kBAAkB,EAAE;gBAClB,YAAY,EAAE,SAAS,CAAC,YAAY,CAAC,QAAQ;gBAC7C,QAAQ,EAAE,SAAS,CAAC,QAAQ,CAAC,QAAQ;gBACrC,QAAQ,EAAE,SAAS,CAAC,QAAQ,CAAC,QAAQ;aACtC;YACD,2BAA2B;SAC5B,EACD,oFAAoF,CACrF,CAAC;QACF,MAAM,IAAI,KAAK,CACb,8DAA8D,2BAA2B,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CACvG,CAAC;IACJ,CAAC;AACH,CAAC;AAED,MAAM,UAAU,8BAA8B,CAAC,MAK9C;IACC,MAAM,EAAE,SAAS,EAAE,YAAY,EAAE,QAAQ,EAAE,MAAM,EAAE,GAAG,MAAM,CAAC;IAC7D,MAAM,iBAAiB,GAAG,wBAAwB,CAAC,YAAY,CAAC,CAAC;IAEjE,IAAI,UAAU,GAAG,QAAQ,CAAC,UAAU,CAAC;IACrC,IAAI,UAAU,GAAG,QAAQ,CAAC,UAAU,CAAC;IACrC,IAAI,mBAAmB,GAAG,QAAQ,CAAC,mBAAmB,CAAC;IAEvD,MAAM,cAAc,GAClB,CAAC,UAAU;QACX,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;QACpC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ,CAAC;IAC3C,MAAM,cAAc,GAClB,CAAC,UAAU;QACX,SAAS,CAAC,QAAQ,CAAC,OAAO,KAAK,KAAK;QACpC,SAAS,CAAC,QAAQ,CAAC,QAAQ,KAAK,QAAQ,CAAC;IAC3C,MAAM,oBAAoB,GACxB,CAAC,mBAAmB;QACpB,SAAS,CAAC,YAAY,CAAC,OAAO,KAAK,KAAK;QACxC,SAAS,CAAC,YAAY,CAAC,QAAQ,KAAK,QAAQ,CAAC;IAE/C,IACE,CAAC,cAAc,IAAI,cAAc,IAAI,oBAAoB,CAAC;QAC1D,CAAC,iBAAiB,CAAC,eAAe;YAChC,iBAAiB,CAAC,eAAe;YACjC,iBAAiB,CAAC,qBAAqB,CAAC,EAC1C,CAAC;QACD,MAAM,CAAC,IAAI,CAAC,oCAAoC,CAAC,CAAC;QAElD,IAAI,cAAc,IAAI,iBAAiB,CAAC,eAAe,EAAE,CAAC;YACxD,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,GAAG,SAAS,EAAE,GAAG,YAAY,EAAE,GAAG,IAAI,EAAE,CAAC;YACrE,UAAU,GAAG,IAAI,SAAS,CACxB;gBACE,MAAM,EAAE,iBAAiB,CAAC,eAAe;gBACzC,GAAG,SAAS;aACb,EACD,MAAM,CACP,CAAC;QACJ,CAAC;QAED,IAAI,cAAc,IAAI,iBAAiB,CAAC,eAAe,EAAE,CAAC;YACxD,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,GAAG,SAAS,EAAE,GAAG,YAAY,EAAE,GAAG,IAAI,EAAE,CAAC;YACrE,UAAU,GAAG,IAAI,SAAS,CACxB;gBACE,MAAM,EAAE,iBAAiB,CAAC,eAAe;gBACzC,KAAK,EAAE,OAAO;gBACd,KAAK,EAAE,wBAAwB;gBAC/B,cAAc,EAAE,KAAK;gBACrB,GAAG,SAAS;aACb,EACD,MAAM,CACP,CAAC;QACJ,CAAC;QAED,MAAM,eAAe,GAAG,iBAAiB,CAAC,qBAAqB,CAAC;QAChE,IAAI,oBAAoB,IAAI,eAAe,EAAE,CAAC;YAC5C,mBAAmB,GAAG;gBACpB,EAAE,EAAE,QAAQ;gBACZ,aAAa,EAAE,CAAC,EAAE,MAAM,EAAE,aAAa,EAAE,QAAQ,EAAE,MAAM,EAAE,EAAE,EAAE,CAC7D,IAAI,kCAAkC,CAAC;oBACrC,MAAM,EAAE,eAAe;oBACvB,MAAM,EAAE,aAAa;oBACrB,kBAAkB,EAChB,YAAY,EAAE,0BAA0B;wBACxC,2CAA2C;oBAC7C,GAAG,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oBACjC,GAAG,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;oBAC7B,aAAa,EAAE,IAAI;iBACpB,CAAC;aACL,CAAC;QACJ,CAAC;IACH,CAAC;SAAM,IAAI,cAAc,IAAI,cAAc,IAAI,oBAAoB,EAAE,CAAC;QACpE,MAAM,CAAC,IAAI,CAAC,oEAAoE,CAAC,CAAC;IACpF,CAAC;IAED,OAAO;QACL,UAAU;QACV,UAAU;QACV,mBAAmB;KACpB,CAAC;AACJ,CAAC"}

package/dist/src/server/speech/providers/openai/stt.js

@@ -0,0 +1,208 @@
+import { EventEmitter } from "node:events";
+import OpenAI from "openai";
+import { writeFile, unlink } from "fs/promises";
+import { join } from "path";
+import { tmpdir } from "os";
+import { v4 } from "uuid";
+import { inferAudioExtension } from "../../../agent/audio-utils.js";
+function isObject(value) {
+    return typeof value === "object" && value !== null;
+}
+function isLogprobToken(value) {
+    if (!isObject(value)) {
+        return false;
+    }
+    if (typeof value.token !== "string") {
+        return false;
+    }
+    if (typeof value.logprob !== "number") {
+        return false;
+    }
+    if (value.bytes === undefined) {
+        return true;
+    }
+    return Array.isArray(value.bytes) && value.bytes.every((entry) => typeof entry === "number");
+}
+function isLogprobTokenArray(value) {
+    return Array.isArray(value) && value.every((entry) => isLogprobToken(entry));
+}
+export class OpenAISTT {
+    constructor(sttConfig, parentLogger) {
+        this.id = "openai";
+        this.config = sttConfig;
+        this.logger = parentLogger.child({ module: "agent", provider: "openai", component: "stt" });
+        this.openaiClient = new OpenAI({
+            apiKey: sttConfig.apiKey,
+        });
+        this.logger.info({ model: sttConfig.model || "whisper-1" }, "STT (OpenAI Whisper) initialized");
+    }
+    createSession(params) {
+        const emitter = new EventEmitter();
+        const logger = params.logger.child({ provider: "openai", component: "stt-session" });
+        const requiredSampleRate = 24000;
+        let connected = false;
+        let segmentId = v4();
+        let previousSegmentId = null;
+        let pcm16 = Buffer.alloc(0);
+        const transcribeAudio = this.transcribeAudioInternal.bind(this);
+        const convertPCMToWavBuffer = (pcmBuffer) => {
+            const headerSize = 44;
+            const channels = 1;
+            const bitsPerSample = 16;
+            const sampleRate = requiredSampleRate;
+            const wavBuffer = Buffer.alloc(headerSize + pcmBuffer.length);
+            const byteRate = (sampleRate * channels * bitsPerSample) / 8;
+            const blockAlign = (channels * bitsPerSample) / 8;
+            wavBuffer.write("RIFF", 0);
+            wavBuffer.writeUInt32LE(36 + pcmBuffer.length, 4);
+            wavBuffer.write("WAVE", 8);
+            wavBuffer.write("fmt ", 12);
+            wavBuffer.writeUInt32LE(16, 16);
+            wavBuffer.writeUInt16LE(1, 20);
+            wavBuffer.writeUInt16LE(channels, 22);
+            wavBuffer.writeUInt32LE(sampleRate, 24);
+            wavBuffer.writeUInt32LE(byteRate, 28);
+            wavBuffer.writeUInt16LE(blockAlign, 32);
+            wavBuffer.writeUInt16LE(bitsPerSample, 34);
+            wavBuffer.write("data", 36);
+            wavBuffer.writeUInt32LE(pcmBuffer.length, 40);
+            pcmBuffer.copy(wavBuffer, 44);
+            return wavBuffer;
+        };
+        return {
+            requiredSampleRate,
+            async connect() {
+                connected = true;
+            },
+            appendPcm16(chunk) {
+                if (!connected) {
+                    emitter.emit("error", new Error("STT session not connected"));
+                    return;
+                }
+                pcm16 = pcm16.length === 0 ? chunk : Buffer.concat([pcm16, chunk]);
+            },
+            commit() {
+                if (!connected) {
+                    emitter.emit("error", new Error("STT session not connected"));
+                    return;
+                }
+                const committedId = segmentId;
+                const prev = previousSegmentId;
+                emitter.emit("committed", { segmentId: committedId, previousSegmentId: prev });
+                void (async () => {
+                    try {
+                        if (pcm16.length === 0) {
+                            emitter.emit("transcript", {
+                                segmentId: committedId,
+                                transcript: "",
+                                isFinal: true,
+                                language: params.language,
+                                isLowConfidence: true,
+                            });
+                            return;
+                        }
+                        const wav = convertPCMToWavBuffer(pcm16);
+                        const result = await transcribeAudio(wav, "audio/wav", params.language ?? "en", logger);
+                        emitter.emit("transcript", {
+                            segmentId: committedId,
+                            transcript: result.text,
+                            isFinal: true,
+                            language: result.language,
+                            logprobs: result.logprobs,
+                            avgLogprob: result.avgLogprob,
+                            isLowConfidence: result.isLowConfidence,
+                        });
+                    }
+                    catch (err) {
+                        emitter.emit("error", err);
+                    }
+                    finally {
+                        previousSegmentId = committedId;
+                        segmentId = v4();
+                        pcm16 = Buffer.alloc(0);
+                    }
+                })();
+            },
+            clear() {
+                pcm16 = Buffer.alloc(0);
+                segmentId = v4();
+            },
+            close() {
+                connected = false;
+                pcm16 = Buffer.alloc(0);
+            },
+            on(event, handler) {
+                emitter.on(event, handler);
+                return undefined;
+            },
+        };
+    }
+    async transcribeAudioInternal(audioBuffer, format, language, logger) {
+        const startTime = Date.now();
+        let tempFilePath = null;
+        try {
+            const ext = inferAudioExtension(format);
+            tempFilePath = join(tmpdir(), `audio-${v4()}.${ext}`);
+            await writeFile(tempFilePath, audioBuffer);
+            logger.debug({ tempFilePath, bytes: audioBuffer.length }, "Transcribing audio file");
+            const modelToUse = this.config.model ?? "whisper-1";
+            const supportsLogprobs = modelToUse === "gpt-4o-transcribe" || modelToUse === "gpt-4o-mini-transcribe";
+            const includeLogprobs = ["logprobs"];
+            const response = await this.openaiClient.audio.transcriptions.create({
+                file: await import("fs").then((fs) => fs.createReadStream(tempFilePath)),
+                language,
+                model: modelToUse,
+                ...(supportsLogprobs ? { include: includeLogprobs } : {}),
+                response_format: "json",
+            });
+            const duration = Date.now() - startTime;
+            const confidenceThreshold = this.config.confidenceThreshold ?? -3.0;
+            let avgLogprob;
+            let isLowConfidence = false;
+            const logprobs = supportsLogprobs &&
+                isObject(response) &&
+                isLogprobTokenArray(response.logprobs)
+                ? response.logprobs
+                : undefined;
+            if (logprobs && logprobs.length > 0) {
+                const totalLogprob = logprobs.reduce((sum, token) => sum + token.logprob, 0);
+                avgLogprob = totalLogprob / logprobs.length;
+                isLowConfidence = avgLogprob < confidenceThreshold;
+                if (isLowConfidence) {
+                    logger.debug({
+                        avgLogprob,
+                        threshold: confidenceThreshold,
+                        text: response.text,
+                        tokenLogprobs: logprobs.map((t) => `${t.token}:${t.logprob.toFixed(2)}`).join(", "),
+                    }, "Low confidence transcription detected");
+                }
+            }
+            logger.debug({ duration, text: response.text, avgLogprob }, "Transcription complete");
+            return {
+                text: response.text,
+                duration: duration,
+                logprobs: logprobs,
+                avgLogprob: avgLogprob,
+                isLowConfidence: isLowConfidence,
+                language: isObject(response) && typeof response.language === "string"
+                    ? response.language
+                    : undefined,
+            };
+        }
+        catch (error) {
+            logger.error({ err: error }, "Transcription error");
+            throw new Error(`STT transcription failed: ${error.message}`);
+        }
+        finally {
+            if (tempFilePath) {
+                try {
+                    await unlink(tempFilePath);
+                }
+                catch (cleanupError) {
+                    logger.warn({ tempFilePath }, "Failed to clean up temp file");
+                }
+            }
+        }
+    }
+}
+//# sourceMappingURL=stt.js.map

package/dist/src/server/speech/providers/openai/stt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"stt.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/stt.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,YAAY,EAAE,MAAM,aAAa,CAAC;AAE3C,OAAO,MAAM,MAAM,QAAQ,CAAC;AAC5B,OAAO,EAAE,SAAS,EAAE,MAAM,EAAE,MAAM,aAAa,CAAC;AAChD,OAAO,EAAE,IAAI,EAAE,MAAM,MAAM,CAAC;AAC5B,OAAO,EAAE,MAAM,EAAE,MAAM,IAAI,CAAC;AAC5B,OAAO,EAAE,EAAE,EAAE,MAAM,MAAM,CAAC;AAC1B,OAAO,EAAE,mBAAmB,EAAE,MAAM,+BAA+B,CAAC;AAgBpE,SAAS,QAAQ,CAAC,KAAc;IAC9B,OAAO,OAAO,KAAK,KAAK,QAAQ,IAAI,KAAK,KAAK,IAAI,CAAC;AACrD,CAAC;AAED,SAAS,cAAc,CAAC,KAAc;IACpC,IAAI,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QACrB,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IAAI,OAAO,KAAK,CAAC,KAAK,KAAK,QAAQ,EAAE,CAAC;QACpC,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IAAI,OAAO,KAAK,CAAC,OAAO,KAAK,QAAQ,EAAE,CAAC;QACtC,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IAAI,KAAK,CAAC,KAAK,KAAK,SAAS,EAAE,CAAC;QAC9B,OAAO,IAAI,CAAC;IACd,CAAC;IACD,OAAO,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,KAAK,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,OAAO,KAAK,KAAK,QAAQ,CAAC,CAAC;AAC/F,CAAC;AAED,SAAS,mBAAmB,CAAC,KAAc;IACzC,OAAO,KAAK,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,KAAK,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,cAAc,CAAC,KAAK,CAAC,CAAC,CAAC;AAC/E,CAAC;AAED,MAAM,OAAO,SAAS;IAMpB,YAAY,SAAoB,EAAE,YAAyB;QAF3C,OAAE,GAAG,QAAiB,CAAC;QAGrC,IAAI,CAAC,MAAM,GAAG,SAAS,CAAC;QACxB,IAAI,CAAC,MAAM,GAAG,YAAY,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,QAAQ,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC;QAC5F,IAAI,CAAC,YAAY,GAAG,IAAI,MAAM,CAAC;YAC7B,MAAM,EAAE,SAAS,CAAC,MAAM;SACzB,CAAC,CAAC;QACH,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,KAAK,EAAE,SAAS,CAAC,KAAK,IAAI,WAAW,EAAE,EAAE,kCAAkC,CAAC,CAAC;IAClG,CAAC;IAEM,aAAa,CAAC,MAIpB;QACC,MAAM,OAAO,GAAG,IAAI,YAAY,EAAE,CAAC;QACnC,MAAM,MAAM,GAAG,MAAM,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,QAAQ,EAAE,SAAS,EAAE,aAAa,EAAE,CAAC,CAAC;QACrF,MAAM,kBAAkB,GAAG,KAAK,CAAC;QAEjC,IAAI,SAAS,GAAG,KAAK,CAAC;QACtB,IAAI,SAAS,GAAG,EAAE,EAAE,CAAC;QACrB,IAAI,iBAAiB,GAAkB,IAAI,CAAC;QAC5C,IAAI,KAAK,GAAW,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QACpC,MAAM,eAAe,GAAG,IAAI,CAAC,uBAAuB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;QAEhE,MAAM,qBAAqB,GAAG,CAAC,SAAiB,EAAU,EAAE;YAC1D,MAAM,UAAU,GAAG,EAAE,CAAC;YACtB,MAAM,QAAQ,GAAG,CAAC,CAAC;YACnB,MAAM,aAAa,GAAG,EAAE,CAAC;YACzB,MAAM,UAAU,GAAG,kBAAkB,CAAC;YACtC,MAAM,SAAS,GAAG,MAAM,CAAC,KAAK,CAAC,UAAU,GAAG,SAAS,CAAC,MAAM,CAAC,CAAC;YAC9D,MAAM,QAAQ,GAAG,CAAC,UAAU,GAAG,QAAQ,GAAG,aAAa,CAAC,GAAG,CAAC,CAAC;YAC7D,MAAM,UAAU,GAAG,CAAC,QAAQ,GAAG,aAAa,CAAC,GAAG,CAAC,CAAC;YAElD,SAAS,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;YAC3B,SAAS,CAAC,aAAa,CAAC,EAAE,GAAG,SAAS,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;YAClD,SAAS,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;YAC3B,SAAS,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;YAC5B,SAAS,CAAC,aAAa,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC;YAChC,SAAS,CAAC,aAAa,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;YAC/B,SAAS,CAAC,aAAa,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC;YACtC,SAAS,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;YACxC,SAAS,CAAC,aAAa,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC;YACtC,SAAS,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;YACxC,SAAS,CAAC,aAAa,CAAC,aAAa,EAAE,EAAE,CAAC,CAAC;YAC3C,SAAS,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;YAC5B,SAAS,CAAC,aAAa,CAAC,SAAS,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;YAC9C,SAAS,CAAC,IAAI,CAAC,SAAS,EAAE,EAAE,CAAC,CAAC;YAE9B,OAAO,SAAS,CAAC;QACnB,CAAC,CAAC;QAEF,OAAO;YACL,kBAAkB;YAClB,KAAK,CAAC,OAAO;gBACX,SAAS,GAAG,IAAI,CAAC;YACnB,CAAC;YACD,WAAW,CAAC,KAAa;gBACvB,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBACD,KAAK,GAAG,KAAK,CAAC,MAAM,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,KAAK,EAAE,KAAK,CAAC,CAAC,CAAC;YACrE,CAAC;YACD,MAAM;gBACJ,IAAI,CAAC,SAAS,EAAE,CAAC;oBACd,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,IAAI,KAAK,CAAC,2BAA2B,CAAC,CAAC,CAAC;oBACvE,OAAO;gBACT,CAAC;gBAED,MAAM,WAAW,GAAG,SAAS,CAAC;gBAC9B,MAAM,IAAI,GAAG,iBAAiB,CAAC;gBAC9B,OAAe,CAAC,IAAI,CAAC,WAAW,EAAE,EAAE,SAAS,EAAE,WAAW,EAAE,iBAAiB,EAAE,IAAI,EAAE,CAAC,CAAC;gBAExF,KAAK,CAAC,KAAK,IAAI,EAAE;oBACf,IAAI,CAAC;wBACH,IAAI,KAAK,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;4BACtB,OAAe,CAAC,IAAI,CAAC,YAAY,EAAE;gCAClC,SAAS,EAAE,WAAW;gCACtB,UAAU,EAAE,EAAE;gCACd,OAAO,EAAE,IAAI;gCACb,QAAQ,EAAE,MAAM,CAAC,QAAQ;gCACzB,eAAe,EAAE,IAAI;6BACtB,CAAC,CAAC;4BACH,OAAO;wBACT,CAAC;wBAED,MAAM,GAAG,GAAG,qBAAqB,CAAC,KAAK,CAAC,CAAC;wBACzC,MAAM,MAAM,GAAG,MAAM,eAAe,CAClC,GAAG,EACH,WAAW,EACX,MAAM,CAAC,QAAQ,IAAI,IAAI,EACvB,MAAM,CACP,CAAC;wBAED,OAAe,CAAC,IAAI,CAAC,YAAY,EAAE;4BAClC,SAAS,EAAE,WAAW;4BACtB,UAAU,EAAE,MAAM,CAAC,IAAI;4BACvB,OAAO,EAAE,IAAI;4BACb,QAAQ,EAAE,MAAM,CAAC,QAAQ;4BACzB,QAAQ,EAAE,MAAM,CAAC,QAAQ;4BACzB,UAAU,EAAE,MAAM,CAAC,UAAU;4BAC7B,eAAe,EAAE,MAAM,CAAC,eAAe;yBACxC,CAAC,CAAC;oBACL,CAAC;oBAAC,OAAO,GAAG,EAAE,CAAC;wBACZ,OAAe,CAAC,IAAI,CAAC,OAAO,EAAE,GAAG,CAAC,CAAC;oBACtC,CAAC;4BAAS,CAAC;wBACT,iBAAiB,GAAG,WAAW,CAAC;wBAChC,SAAS,GAAG,EAAE,EAAE,CAAC;wBACjB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;oBAC1B,CAAC;gBACH,CAAC,CAAC,EAAE,CAAC;YACP,CAAC;YACD,KAAK;gBACH,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;gBACxB,SAAS,GAAG,EAAE,EAAE,CAAC;YACnB,CAAC;YACD,KAAK;gBACH,SAAS,GAAG,KAAK,CAAC;gBAClB,KAAK,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;YAC1B,CAAC;YACD,EAAE,CAAC,KAAU,EAAE,OAAY;gBACzB,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,OAAO,CAAC,CAAC;gBAC3B,OAAO,SAAS,CAAC;YACnB,CAAC;SACF,CAAC;IACJ,CAAC;IAEO,KAAK,CAAC,uBAAuB,CACnC,WAAmB,EACnB,MAAc,EACd,QAAgB,EAChB,MAAmB;QAEnB,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAC7B,IAAI,YAAY,GAAkB,IAAI,CAAC;QAEvC,IAAI,CAAC;YACH,MAAM,GAAG,GAAG,mBAAmB,CAAC,MAAM,CAAC,CAAC;YACxC,YAAY,GAAG,IAAI,CAAC,MAAM,EAAE,EAAE,SAAS,EAAE,EAAE,IAAI,GAAG,EAAE,CAAC,CAAC;YACtD,MAAM,SAAS,CAAC,YAAY,EAAE,WAAW,CAAC,CAAC;YAE3C,MAAM,CAAC,KAAK,CACV,EAAE,YAAY,EAAE,KAAK,EAAE,WAAW,CAAC,MAAM,EAAE,EAC3C,yBAAyB,CAC1B,CAAC;YAEF,MAAM,UAAU,GAAG,IAAI,CAAC,MAAM,CAAC,KAAK,IAAI,WAAW,CAAC;YACpD,MAAM,gBAAgB,GACpB,UAAU,KAAK,mBAAmB,IAAI,UAAU,KAAK,wBAAwB,CAAC;YAChF,MAAM,eAAe,GAAiB,CAAC,UAAU,CAAC,CAAC;YAEnD,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,KAAK,CAAC,cAAc,CAAC,MAAM,CAAC;gBACnE,IAAI,EAAE,MAAM,MAAM,CAAC,IAAI,CAAC,CAAC,IAAI,CAAC,CAAC,EAAE,EAAE,EAAE,CAAC,EAAE,CAAC,gBAAgB,CAAC,YAAa,CAAC,CAAC;gBACzE,QAAQ;gBACR,KAAK,EAAE,UAAU;gBACjB,GAAG,CAAC,gBAAgB,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,eAAe,EAAE,CAAC,CAAC,CAAC,EAAE,CAAC;gBACzD,eAAe,EAAE,MAAM;aACxB,CAAC,CAAC;YAEH,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS,CAAC;YACxC,MAAM,mBAAmB,GAAG,IAAI,CAAC,MAAM,CAAC,mBAAmB,IAAI,CAAC,GAAG,CAAC;YAEpE,IAAI,UAA8B,CAAC;YACnC,IAAI,eAAe,GAAG,KAAK,CAAC;YAC5B,MAAM,QAAQ,GACZ,gBAAgB;gBAChB,QAAQ,CAAC,QAAQ,CAAC;gBAClB,mBAAmB,CAAC,QAAQ,CAAC,QAAQ,CAAC;gBACpC,CAAC,CAAC,QAAQ,CAAC,QAAQ;gBACnB,CAAC,CAAC,SAAS,CAAC;YAEhB,IAAI,QAAQ,IAAI,QAAQ,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;gBACpC,MAAM,YAAY,GAAG,QAAQ,CAAC,MAAM,CAAC,CAAC,GAAG,EAAE,KAAK,EAAE,EAAE,CAAC,GAAG,GAAG,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC;gBAC7E,UAAU,GAAG,YAAY,GAAG,QAAQ,CAAC,MAAM,CAAC;gBAC5C,eAAe,GAAG,UAAU,GAAG,mBAAmB,CAAC;gBAEnD,IAAI,eAAe,EAAE,CAAC;oBACpB,MAAM,CAAC,KAAK,CACV;wBACE,UAAU;wBACV,SAAS,EAAE,mBAAmB;wBAC9B,IAAI,EAAE,QAAQ,CAAC,IAAI;wBACnB,aAAa,EAAE,QAAQ,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,GAAG,CAAC,CAAC,KAAK,IAAI,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC;qBACpF,EACD,uCAAuC,CACxC,CAAC;gBACJ,CAAC;YACH,CAAC;YAED,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,IAAI,EAAE,QAAQ,CAAC,IAAI,EAAE,UAAU,EAAE,EAAE,wBAAwB,CAAC,CAAC;YAEtF,OAAO;gBACL,IAAI,EAAE,QAAQ,CAAC,IAAI;gBACnB,QAAQ,EAAE,QAAQ;gBAClB,QAAQ,EAAE,QAAQ;gBAClB,UAAU,EAAE,UAAU;gBACtB,eAAe,EAAE,eAAe;gBAChC,QAAQ,EACN,QAAQ,CAAC,QAAQ,CAAC,IAAI,OAAO,QAAQ,CAAC,QAAQ,KAAK,QAAQ;oBACzD,CAAC,CAAC,QAAQ,CAAC,QAAQ;oBACnB,CAAC,CAAC,SAAS;aAChB,CAAC;QACJ,CAAC;QAAC,OAAO,KAAU,EAAE,CAAC;YACpB,MAAM,CAAC,KAAK,CAAC,EAAE,GAAG,EAAE,KAAK,EAAE,EAAE,qBAAqB,CAAC,CAAC;YACpD,MAAM,IAAI,KAAK,CAAC,6BAA6B,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;QAChE,CAAC;gBAAS,CAAC;YACT,IAAI,YAAY,EAAE,CAAC;gBACjB,IAAI,CAAC;oBACH,MAAM,MAAM,CAAC,YAAY,CAAC,CAAC;gBAC7B,CAAC;gBAAC,OAAO,YAAY,EAAE,CAAC;oBACtB,MAAM,CAAC,IAAI,CAAC,EAAE,YAAY,EAAE,EAAE,8BAA8B,CAAC,CAAC;gBAChE,CAAC;YACH,CAAC;QACH,CAAC;IACH,CAAC;CACF"}

package/dist/src/server/speech/providers/openai/tts.js

@@ -0,0 +1,46 @@
+import OpenAI from "openai";
+export class OpenAITTS {
+    constructor(ttsConfig, parentLogger) {
+        this.config = {
+            model: "tts-1",
+            voice: "alloy",
+            responseFormat: "pcm",
+            ...ttsConfig,
+        };
+        this.logger = parentLogger.child({ module: "agent", provider: "openai", component: "tts" });
+        this.openaiClient = new OpenAI({
+            apiKey: ttsConfig.apiKey,
+        });
+        this.logger.info({ voice: this.config.voice, model: this.config.model, format: this.config.responseFormat }, "TTS (OpenAI) initialized");
+    }
+    getConfig() {
+        return this.config;
+    }
+    async synthesizeSpeech(text) {
+        if (!text || text.trim().length === 0) {
+            throw new Error("Cannot synthesize empty text");
+        }
+        const startTime = Date.now();
+        try {
+            this.logger.debug({ textLength: text.length, preview: text.substring(0, 50) }, "Synthesizing speech");
+            const response = await this.openaiClient.audio.speech.create({
+                model: this.config.model,
+                voice: this.config.voice,
+                input: text,
+                response_format: this.config.responseFormat,
+            });
+            const audioStream = response.body;
+            const duration = Date.now() - startTime;
+            this.logger.debug({ duration }, "Speech synthesis stream ready");
+            return {
+                stream: audioStream,
+                format: this.config.responseFormat || "mp3",
+            };
+        }
+        catch (error) {
+            this.logger.error({ err: error }, "Speech synthesis error");
+            throw new Error(`TTS synthesis failed: ${error.message}`);
+        }
+    }
+}
+//# sourceMappingURL=tts.js.map

package/dist/src/server/speech/providers/openai/tts.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"tts.js","sourceRoot":"","sources":["../../../../../../src/server/speech/providers/openai/tts.ts"],"names":[],"mappings":"AACA,OAAO,MAAM,MAAM,QAAQ,CAAC;AAa5B,MAAM,OAAO,SAAS;IAKpB,YAAY,SAAoB,EAAE,YAAyB;QACzD,IAAI,CAAC,MAAM,GAAG;YACZ,KAAK,EAAE,OAAO;YACd,KAAK,EAAE,OAAO;YACd,cAAc,EAAE,KAAK;YACrB,GAAG,SAAS;SACb,CAAC;QACF,IAAI,CAAC,MAAM,GAAG,YAAY,CAAC,KAAK,CAAC,EAAE,MAAM,EAAE,OAAO,EAAE,QAAQ,EAAE,QAAQ,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC,CAAC;QAC5F,IAAI,CAAC,YAAY,GAAG,IAAI,MAAM,CAAC;YAC7B,MAAM,EAAE,SAAS,CAAC,MAAM;SACzB,CAAC,CAAC;QAEH,IAAI,CAAC,MAAM,CAAC,IAAI,CACd,EAAE,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,cAAc,EAAE,EAC1F,0BAA0B,CAC3B,CAAC;IACJ,CAAC;IAEM,SAAS;QACd,OAAO,IAAI,CAAC,MAAM,CAAC;IACrB,CAAC;IAEM,KAAK,CAAC,gBAAgB,CAAC,IAAY;QACxC,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;YACtC,MAAM,IAAI,KAAK,CAAC,8BAA8B,CAAC,CAAC;QAClD,CAAC;QAED,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAE7B,IAAI,CAAC;YACH,IAAI,CAAC,MAAM,CAAC,KAAK,CACf,EAAE,UAAU,EAAE,IAAI,CAAC,MAAM,EAAE,OAAO,EAAE,IAAI,CAAC,SAAS,CAAC,CAAC,EAAE,EAAE,CAAC,EAAE,EAC3D,qBAAqB,CACtB,CAAC;YAEF,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,YAAY,CAAC,KAAK,CAAC,MAAM,CAAC,MAAM,CAAC;gBAC3D,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,KAAM;gBACzB,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,KAAM;gBACzB,KAAK,EAAE,IAAI;gBACX,eAAe,EAAE,IAAI,CAAC,MAAM,CAAC,cAAiE;aAC/F,CAAC,CAAC;YAEH,MAAM,WAAW,GAAG,QAAQ,CAAC,IAA2B,CAAC;YAEzD,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS,CAAC;YACxC,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,QAAQ,EAAE,EAAE,+BAA+B,CAAC,CAAC;YAEjE,OAAO;gBACL,MAAM,EAAE,WAAW;gBACnB,MAAM,EAAE,IAAI,CAAC,MAAM,CAAC,cAAc,IAAI,KAAK;aAC5C,CAAC;QACJ,CAAC;QAAC,OAAO,KAAU,EAAE,CAAC;YACpB,IAAI,CAAC,MAAM,CAAC,KAAK,CAAC,EAAE,GAAG,EAAE,KAAK,EAAE,EAAE,wBAAwB,CAAC,CAAC;YAC5D,MAAM,IAAI,KAAK,CAAC,yBAAyB,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;QAC5D,CAAC;IACH,CAAC;CACF"}

package/dist/src/server/speech/speech-config-resolver.js

@@ -0,0 +1,85 @@
+import { z } from "zod";
+import { resolveLocalSpeechConfig } from "./providers/local/config.js";
+import { resolveOpenAiSpeechConfig } from "./providers/openai/config.js";
+import { SpeechProviderIdSchema, } from "./speech-types.js";
+const OptionalSpeechProviderSchema = z
+    .string()
+    .trim()
+    .toLowerCase()
+    .pipe(SpeechProviderIdSchema)
+    .optional();
+const OptionalBooleanFlagSchema = z
+    .union([z.boolean(), z.string().trim().toLowerCase()])
+    .optional()
+    .transform((value) => {
+    if (typeof value === "boolean") {
+        return value;
+    }
+    if (value === undefined) {
+        return undefined;
+    }
+    if (["1", "true", "yes", "y", "on"].includes(value)) {
+        return true;
+    }
+    if (["0", "false", "no", "n", "off"].includes(value)) {
+        return false;
+    }
+    return undefined;
+});
+const RequestedSpeechProvidersSchema = z.object({
+    dictationStt: OptionalSpeechProviderSchema.default("local"),
+    voiceStt: OptionalSpeechProviderSchema.default("local"),
+    voiceTts: OptionalSpeechProviderSchema.default("local"),
+});
+function resolveRequestedSpeechProviders(params) {
+    const resolveFeatureProvider = (configuredValue, parsedValue, enabled) => ({
+        provider: parsedValue,
+        explicit: configuredValue !== undefined,
+        enabled,
+    });
+    const dictationSttProviderFromConfig = params.env.PASEO_DICTATION_STT_PROVIDER ??
+        params.persisted.features?.dictation?.stt?.provider;
+    const voiceSttProviderFromConfig = params.env.PASEO_VOICE_STT_PROVIDER ??
+        params.persisted.features?.voiceMode?.stt?.provider;
+    const voiceTtsProviderFromConfig = params.env.PASEO_VOICE_TTS_PROVIDER ??
+        params.persisted.features?.voiceMode?.tts?.provider;
+    const dictationEnabled = OptionalBooleanFlagSchema.parse(params.env.PASEO_DICTATION_ENABLED ?? params.persisted.features?.dictation?.enabled) ?? true;
+    const voiceModeEnabled = OptionalBooleanFlagSchema.parse(params.env.PASEO_VOICE_MODE_ENABLED ?? params.persisted.features?.voiceMode?.enabled) ?? true;
+    const parsed = RequestedSpeechProvidersSchema.parse({
+        dictationStt: dictationSttProviderFromConfig ?? "local",
+        voiceStt: voiceSttProviderFromConfig ?? "local",
+        voiceTts: voiceTtsProviderFromConfig ?? "local",
+    });
+    return {
+        dictationStt: resolveFeatureProvider(dictationSttProviderFromConfig, parsed.dictationStt, dictationEnabled),
+        voiceStt: resolveFeatureProvider(voiceSttProviderFromConfig, parsed.voiceStt, voiceModeEnabled),
+        voiceTts: resolveFeatureProvider(voiceTtsProviderFromConfig, parsed.voiceTts, voiceModeEnabled),
+    };
+}
+export function resolveSpeechConfig(params) {
+    const providers = resolveRequestedSpeechProviders({
+        env: params.env,
+        persisted: params.persisted,
+    });
+    const local = resolveLocalSpeechConfig({
+        paseoHome: params.paseoHome,
+        env: params.env,
+        persisted: params.persisted,
+        providers,
+    });
+    const openai = resolveOpenAiSpeechConfig({
+        env: params.env,
+        persisted: params.persisted,
+        providers,
+    });
+    return {
+        openai,
+        speech: {
+            providers,
+            ...(local.local
+                ? { local: local.local }
+                : {}),
+        },
+    };
+}
+//# sourceMappingURL=speech-config-resolver.js.map

package/dist/src/server/speech/speech-config-resolver.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"speech-config-resolver.js","sourceRoot":"","sources":["../../../../src/server/speech/speech-config-resolver.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,CAAC,EAAE,MAAM,KAAK,CAAC;AAIxB,OAAO,EAAE,wBAAwB,EAAE,MAAM,6BAA6B,CAAC;AACvE,OAAO,EAAE,yBAAyB,EAAE,MAAM,8BAA8B,CAAC;AACzE,OAAO,EACL,sBAAsB,GAGvB,MAAM,mBAAmB,CAAC;AAE3B,MAAM,4BAA4B,GAAG,CAAC;KACnC,MAAM,EAAE;KACR,IAAI,EAAE;KACN,WAAW,EAAE;KACb,IAAI,CAAC,sBAAsB,CAAC;KAC5B,QAAQ,EAAE,CAAC;AAEd,MAAM,yBAAyB,GAAG,CAAC;KAChC,KAAK,CAAC,CAAC,CAAC,CAAC,OAAO,EAAE,EAAE,CAAC,CAAC,MAAM,EAAE,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC,CAAC;KACrD,QAAQ,EAAE;KACV,SAAS,CAAC,CAAC,KAAK,EAAE,EAAE;IACnB,IAAI,OAAO,KAAK,KAAK,SAAS,EAAE,CAAC;QAC/B,OAAO,KAAK,CAAC;IACf,CAAC;IACD,IAAI,KAAK,KAAK,SAAS,EAAE,CAAC;QACxB,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,IAAI,CAAC,GAAG,EAAE,MAAM,EAAE,KAAK,EAAE,GAAG,EAAE,IAAI,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QACpD,OAAO,IAAI,CAAC;IACd,CAAC;IACD,IAAI,CAAC,GAAG,EAAE,OAAO,EAAE,IAAI,EAAE,GAAG,EAAE,KAAK,CAAC,CAAC,QAAQ,CAAC,KAAK,CAAC,EAAE,CAAC;QACrD,OAAO,KAAK,CAAC;IACf,CAAC;IACD,OAAO,SAAS,CAAC;AACnB,CAAC,CAAC,CAAC;AAEL,MAAM,8BAA8B,GAAG,CAAC,CAAC,MAAM,CAAC;IAC9C,YAAY,EAAE,4BAA4B,CAAC,OAAO,CAAC,OAAO,CAAC;IAC3D,QAAQ,EAAE,4BAA4B,CAAC,OAAO,CAAC,OAAO,CAAC;IACvD,QAAQ,EAAE,4BAA4B,CAAC,OAAO,CAAC,OAAO,CAAC;CACxD,CAAC,CAAC;AAEH,SAAS,+BAA+B,CAAC,MAGxC;IACC,MAAM,sBAAsB,GAAG,CAC7B,eAAmC,EACnC,WAAmD,EACnD,OAAgB,EACS,EAAE,CAAC,CAAC;QAC7B,QAAQ,EAAE,WAAW;QACrB,QAAQ,EAAE,eAAe,KAAK,SAAS;QACvC,OAAO;KACR,CAAC,CAAC;IAEH,MAAM,8BAA8B,GAClC,MAAM,CAAC,GAAG,CAAC,4BAA4B;QACvC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,QAAQ,CAAC;IACtD,MAAM,0BAA0B,GAC9B,MAAM,CAAC,GAAG,CAAC,wBAAwB;QACnC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,QAAQ,CAAC;IACtD,MAAM,0BAA0B,GAC9B,MAAM,CAAC,GAAG,CAAC,wBAAwB;QACnC,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,GAAG,EAAE,QAAQ,CAAC;IACtD,MAAM,gBAAgB,GACpB,yBAAyB,CAAC,KAAK,CAC7B,MAAM,CAAC,GAAG,CAAC,uBAAuB,IAAI,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,OAAO,CACpF,IAAI,IAAI,CAAC;IACZ,MAAM,gBAAgB,GACpB,yBAAyB,CAAC,KAAK,CAC7B,MAAM,CAAC,GAAG,CAAC,wBAAwB,IAAI,MAAM,CAAC,SAAS,CAAC,QAAQ,EAAE,SAAS,EAAE,OAAO,CACrF,IAAI,IAAI,CAAC;IAEZ,MAAM,MAAM,GAAG,8BAA8B,CAAC,KAAK,CAAC;QAClD,YAAY,EAAE,8BAA8B,IAAI,OAAO;QACvD,QAAQ,EAAE,0BAA0B,IAAI,OAAO;QAC/C,QAAQ,EAAE,0BAA0B,IAAI,OAAO;KAChD,CAAC,CAAC;IAEH,OAAO;QACL,YAAY,EAAE,sBAAsB,CAClC,8BAA8B,EAC9B,MAAM,CAAC,YAAY,EACnB,gBAAgB,CACjB;QACD,QAAQ,EAAE,sBAAsB,CAC9B,0BAA0B,EAC1B,MAAM,CAAC,QAAQ,EACf,gBAAgB,CACjB;QACD,QAAQ,EAAE,sBAAsB,CAC9B,0BAA0B,EAC1B,MAAM,CAAC,QAAQ,EACf,gBAAgB,CACjB;KACF,CAAC;AACJ,CAAC;AAED,MAAM,UAAU,mBAAmB,CAAC,MAInC;IAIC,MAAM,SAAS,GAAG,+BAA+B,CAAC;QAChD,GAAG,EAAE,MAAM,CAAC,GAAG;QACf,SAAS,EAAE,MAAM,CAAC,SAAS;KAC5B,CAAC,CAAC;IAEH,MAAM,KAAK,GAAG,wBAAwB,CAAC;QACrC,SAAS,EAAE,MAAM,CAAC,SAAS;QAC3B,GAAG,EAAE,MAAM,CAAC,GAAG;QACf,SAAS,EAAE,MAAM,CAAC,SAAS;QAC3B,SAAS;KACV,CAAC,CAAC;IAEH,MAAM,MAAM,GAAG,yBAAyB,CAAC;QACvC,GAAG,EAAE,MAAM,CAAC,GAAG;QACf,SAAS,EAAE,MAAM,CAAC,SAAS;QAC3B,SAAS;KACV,CAAC,CAAC;IAEH,OAAO;QACL,MAAM;QACN,MAAM,EAAE;YACN,SAAS;YACT,GAAG,CAAC,KAAK,CAAC,KAAK;gBACb,CAAC,CAAC,EAAE,KAAK,EAAE,KAAK,CAAC,KAAK,EAAE;gBACxB,CAAC,CAAC,EAAE,CAAC;SACR;KACF,CAAC;AACJ,CAAC"}

package/dist/src/server/speech/speech-provider.js

@@ -0,0 +1,2 @@
+export {};
+//# sourceMappingURL=speech-provider.js.map

package/dist/src/server/speech/speech-provider.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"speech-provider.js","sourceRoot":"","sources":["../../../../src/server/speech/speech-provider.ts"],"names":[],"mappings":""}