chat-flow-ardymalihi 1.0.2 → 3.0.0

package/dist/adapters/voice/GeminiMultimodalLiveAdapter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"GeminiMultimodalLiveAdapter.js","sourceRoot":"","sources":["../../../src/adapters/voice/GeminiMultimodalLiveAdapter.ts"],"names":[],"mappings":";;;AAEA;;;;;;;;;;;;;;;;;GAiBG;AACH,MAAa,2BAA2B;IAMpC,YAAY,MAAc,EAAE,WAAyB;QAF7C,gBAAW,GAAY,KAAK,CAAC;QAGjC,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,WAAW,GAAG,WAAW,CAAC;IACnC,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,OAAO;QACT,IAAI,IAAI,CAAC,EAAE,IAAI,IAAI,CAAC,WAAW,EAAE,CAAC;YAC9B,OAAO,IAAI,CAAC,EAAE,CAAC;QACnB,CAAC;QAED,4CAA4C;QAC5C,MAAM,KAAK,GAAG,6HAA6H,IAAI,CAAC,MAAM,EAAE,CAAC;QAEzJ,OAAO,IAAI,OAAO,CAAC,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;YACnC,IAAI,CAAC,EAAE,GAAG,IAAI,SAAS,CAAC,KAAK,CAAC,CAAC;YAE/B,IAAI,CAAC,EAAE,CAAC,MAAM,GAAG,GAAG,EAAE;gBAClB,OAAO,CAAC,GAAG,CAAC,4CAA4C,CAAC,CAAC;gBAC1D,IAAI,CAAC,WAAW,GAAG,IAAI,CAAC;gBAExB,6BAA6B;gBAC7B,IAAI,CAAC,UAAU,EAAE,CAAC;gBAElB,OAAO,CAAC,IAAI,CAAC,EAAG,CAAC,CAAC;YACtB,CAAC,CAAC;YAEF,IAAI,CAAC,EAAE,CAAC,OAAO,GAAG,CAAC,KAAK,EAAE,EAAE;gBACxB,OAAO,CAAC,KAAK,CAAC,0CAA0C,EAAE,KAAK,CAAC,CAAC;gBACjE,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC;gBACzB,MAAM,CAAC,KAAK,CAAC,CAAC;YAClB,CAAC,CAAC;YAEF,IAAI,CAAC,EAAE,CAAC,OAAO,GAAG,GAAG,EAAE;gBACnB,OAAO,CAAC,GAAG,CAAC,0CAA0C,CAAC,CAAC;gBACxD,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC;YAC7B,CAAC,CAAC;QACN,CAAC,CAAC,CAAC;IACP,CAAC;IAED;;OAEG;IACK,UAAU;QACd,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC;YAChC,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAC;QAC/C,CAAC;QAED,MAAM,MAAM,GAAG;YACX,KAAK,EAAE;gBACH,KAAK,EAAE,6BAA6B;gBACpC,gBAAgB,EAAE;oBACd,kBAAkB,EAAE,CAAC,OAAO,CAAC;oBAC7B,YAAY,EAAE;wBACV,WAAW,EAAE;4BACT,mBAAmB,EAAE;gCACjB,SAAS,EAAE,IAAI,CAAC,WAAW,EAAE,GAAG,EAAE,KAAK,IAAI,MAAM;6BACpD;yBACJ;qBACJ;iBACJ;aACJ;SACJ,CAAC;QAEF,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,CAAC;IACzC,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,SAAS,CAAC,WAAmB;QAC/B,IAAI,CAAC,IAAI,CAAC,EAAE,IAAI,CAAC,IAAI,CAAC,WAAW,EAAE,CAAC;YAChC,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAC;QAC/C,CAAC;QAED,MAAM,OAAO,GAAG;YACZ,aAAa,EAAE;gBACX,WAAW,EAAE,CAAC;wBACV,IAAI,EAAE,WAAW;wBACjB,QAAQ,EAAE,WAAW;qBACxB,CAAC;aACL;SACJ,CAAC;QAEF,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC,CAAC,CAAC;IAC1C,CAAC;IAED;;;OAGG;IACH,UAAU,CAAC,QAID;QACN,IAAI,CAAC,IAAI,CAAC,EAAE,EAAE,CAAC;YACX,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAC;QAC/C,CAAC;QAED,IAAI,CAAC,EAAE,CAAC,SAAS,GAAG,CAAC,KAAK,EAAE,EAAE;YAC1B,IAAI,CAAC;gBACD,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,IAAI,CAAC,CAAC;gBAEpC,IAAI,IAAI,CAAC,aAAa,EAAE,SAAS,EAAE,KAAK,EAAE,CAAC;oBACvC,MAAM,KAAK,GAAG,IAAI,CAAC,aAAa,CAAC,SAAS,CAAC,KAAK,CAAC;oBAEjD,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;wBACvB,MAAM,QAAQ,GAAQ;4BAClB,OAAO,EAAE,IAAI,CAAC,aAAa,CAAC,YAAY,IAAI,KAAK;yBACpD,CAAC;wBAEF,eAAe;wBACf,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC;4BACZ,QAAQ,CAAC,IAAI,GAAG,IAAI,CAAC,IAAI,CAAC;wBAC9B,CAAC;wBAED,gBAAgB;wBAChB,IAAI,IAAI,CAAC,UAAU,EAAE,IAAI,EAAE,CAAC;4BACxB,MAAM,WAAW,GAAG,IAAI,CAAC,UAAU,CAAC,IAAI,CAAC;4BACzC,QAAQ,CAAC,KAAK,GAAG,UAAU,CAAC,IAAI,CAC5B,MAAM,CAAC,IAAI,CAAC,WAAW,EAAE,QAAQ,CAAC,CACrC,CAAC;wBACN,CAAC;wBAED,QAAQ,CAAC,QAAQ,CAAC,CAAC;oBACvB,CAAC;gBACL,CAAC;YACL,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBACb,OAAO,CAAC,KAAK,CAAC,gDAAgD,EAAE,KAAK,CAAC,CAAC;YAC3E,CAAC;QACL,CAAC,CAAC;IACN,CAAC;IAED;;OAEG;IACH,UAAU;QACN,IAAI,IAAI,CAAC,EAAE,EAAE,CAAC;YACV,IAAI,CAAC,EAAE,CAAC,KAAK,EAAE,CAAC;YAChB,IAAI,CAAC,EAAE,GAAG,SAAS,CAAC;YACpB,IAAI,CAAC,WAAW,GAAG,KAAK,CAAC;QAC7B,CAAC;IACL,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,YAAY,CAAC,WAAmB,EAAE,MAA2B;QAC/D,MAAM,IAAI,KAAK,CACX,qDAAqD;YACrD,2DAA2D,CAC9D,CAAC;IACN,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,YAAY,CAAC,IAAY,EAAE,MAA2B;QACxD,MAAM,IAAI,KAAK,CACX,qDAAqD;YACrD,2DAA2D,CAC9D,CAAC;IACN,CAAC;IAED;;;OAGG;IACH,kBAAkB;QACd,OAAO;YACH,MAAM,EAAE,KAAK;YACb,UAAU,EAAE,KAAK;YACjB,QAAQ,EAAE,CAAC;YACX,QAAQ,EAAE,EAAE;SACf,CAAC;IACN,CAAC;IAED;;OAEG;IACH,mBAAmB;QACf,OAAO,IAAI,CAAC,WAAW,CAAC;IAC5B,CAAC;IAED;;OAEG;IACH,YAAY;QACR,OAAO,IAAI,CAAC,EAAE,CAAC;IACnB,CAAC;CACJ;AA3MD,kEA2MC"}

package/dist/adapters/voice/GeminiVoiceAdapter.d.ts

@@ -0,0 +1,36 @@
+import { IVoiceAdapter, VoiceConfig, AudioFormatMetadata } from '../../interfaces';
+/**
+ * Gemini Voice Adapter
+ * Implements voice capabilities using Google's Gemini API
+ * Can be used independently of GeminiLLMAdapter
+ */
+export declare class GeminiVoiceAdapter implements IVoiceAdapter {
+    private apiKey;
+    private voiceConfig?;
+    constructor(apiKey: string, voiceConfig?: VoiceConfig);
+    /**
+     * Convert speech (audio) to text using Gemini's audio input capability
+     * @param audioBase64 Base64-encoded audio data (WAV format recommended)
+     * @param config Optional STT configuration (overrides constructor config)
+     * @returns Transcribed text
+     */
+    speechToText(audioBase64: string, config?: VoiceConfig['stt']): Promise<string>;
+    /**
+     * Convert text to speech (audio) using Gemini's native audio output
+     * @param text Text to convert to speech
+     * @param config Optional TTS configuration (overrides constructor config)
+     * @returns Audio data as Uint8Array (PCM format)
+     */
+    textToSpeech(text: string, config?: VoiceConfig['tts']): Promise<Uint8Array>;
+    /**
+     * Declare the audio format produced by Gemini TTS
+     * Used by IAudioStorage for format conversion
+     */
+    getTTSOutputFormat(): AudioFormatMetadata;
+    /**
+     * Declare the expected audio format for Gemini STT
+     * Used by IAudioStorage for format conversion
+     */
+    getSTTInputFormat(): AudioFormatMetadata;
+}
+//# sourceMappingURL=GeminiVoiceAdapter.d.ts.map

package/dist/adapters/voice/GeminiVoiceAdapter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"GeminiVoiceAdapter.d.ts","sourceRoot":"","sources":["../../../src/adapters/voice/GeminiVoiceAdapter.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,aAAa,EAAE,WAAW,EAAE,mBAAmB,EAAE,MAAM,kBAAkB,CAAC;AAEnF;;;;GAIG;AACH,qBAAa,kBAAmB,YAAW,aAAa;IACpD,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,WAAW,CAAC,CAAc;gBAEtB,MAAM,EAAE,MAAM,EAAE,WAAW,CAAC,EAAE,WAAW;IAKrD;;;;;OAKG;IACG,YAAY,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,CAAC,KAAK,CAAC,GAAG,OAAO,CAAC,MAAM,CAAC;IAiDrF;;;;;OAKG;IACG,YAAY,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,CAAC,KAAK,CAAC,GAAG,OAAO,CAAC,UAAU,CAAC;IAiDlF;;;OAGG;IACH,kBAAkB,IAAI,mBAAmB;IASzC;;;OAGG;IACH,iBAAiB,IAAI,mBAAmB;CAQ3C"}

package/dist/adapters/voice/GeminiVoiceAdapter.js

@@ -0,0 +1,134 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.GeminiVoiceAdapter = void 0;
+/**
+ * Gemini Voice Adapter
+ * Implements voice capabilities using Google's Gemini API
+ * Can be used independently of GeminiLLMAdapter
+ */
+class GeminiVoiceAdapter {
+    constructor(apiKey, voiceConfig) {
+        this.apiKey = apiKey;
+        this.voiceConfig = voiceConfig;
+    }
+    /**
+     * Convert speech (audio) to text using Gemini's audio input capability
+     * @param audioBase64 Base64-encoded audio data (WAV format recommended)
+     * @param config Optional STT configuration (overrides constructor config)
+     * @returns Transcribed text
+     */
+    async speechToText(audioBase64, config) {
+        const sttConfig = config || this.voiceConfig?.stt;
+        const language = sttConfig?.language || 'en-US';
+        try {
+            const response = await fetch(`https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-lite:generateContent?key=${this.apiKey}`, {
+                method: 'POST',
+                headers: { 'Content-Type': 'application/json' },
+                body: JSON.stringify({
+                    contents: [{
+                            role: 'user',
+                            parts: [
+                                { text: "Transcribe the following audio exactly. Do not respond to the audio, just output the transcription." },
+                                {
+                                    inlineData: {
+                                        mimeType: 'audio/wav',
+                                        data: audioBase64
+                                    }
+                                }
+                            ]
+                        }],
+                    generationConfig: {
+                        temperature: 0.1, // Lower temperature for transcription accuracy
+                    }
+                })
+            });
+            if (!response.ok) {
+                const errorText = await response.text();
+                throw new Error(`Gemini STT API Error: ${response.status} - ${errorText}`);
+            }
+            const json = await response.json();
+            const transcript = json.candidates?.[0]?.content?.parts?.[0]?.text || '';
+            if (!transcript) {
+                throw new Error('No transcript returned from Gemini STT');
+            }
+            return transcript;
+        }
+        catch (error) {
+            console.error('Speech-to-Text Error:', error);
+            throw new Error(`STT failed: ${error.message}`);
+        }
+    }
+    /**
+     * Convert text to speech (audio) using Gemini's native audio output
+     * @param text Text to convert to speech
+     * @param config Optional TTS configuration (overrides constructor config)
+     * @returns Audio data as Uint8Array (PCM format)
+     */
+    async textToSpeech(text, config) {
+        const ttsConfig = config || this.voiceConfig?.tts;
+        const voice = ttsConfig?.voice || 'Puck'; // Gemini's default voice
+        try {
+            // Use v1alpha API which supports audio modalities
+            const response = await fetch(`https://generativelanguage.googleapis.com/v1alpha/models/gemini-2.0-flash-exp:generateContent?key=${this.apiKey}`, {
+                method: 'POST',
+                headers: { 'Content-Type': 'application/json' },
+                body: JSON.stringify({
+                    contents: [{
+                            role: 'user',
+                            parts: [{ text }]
+                        }],
+                    generationConfig: {
+                        responseModalities: ['AUDIO'],
+                        speechConfig: {
+                            voiceConfig: {
+                                prebuiltVoiceConfig: {
+                                    voiceName: voice
+                                }
+                            }
+                        }
+                    }
+                })
+            });
+            if (!response.ok) {
+                const errorText = await response.text();
+                throw new Error(`Gemini TTS API Error: ${response.status} - ${errorText}`);
+            }
+            const json = await response.json();
+            const audioBase64 = json.candidates?.[0]?.content?.parts?.[0]?.inlineData?.data;
+            if (!audioBase64) {
+                throw new Error('No audio data received from Gemini TTS');
+            }
+            return Uint8Array.from(Buffer.from(audioBase64, 'base64'));
+        }
+        catch (error) {
+            console.error('Text-to-Speech Error:', error);
+            throw new Error(`TTS failed: ${error.message}`);
+        }
+    }
+    /**
+     * Declare the audio format produced by Gemini TTS
+     * Used by IAudioStorage for format conversion
+     */
+    getTTSOutputFormat() {
+        return {
+            format: 'pcm',
+            sampleRate: 24000,
+            channels: 1,
+            bitDepth: 16
+        };
+    }
+    /**
+     * Declare the expected audio format for Gemini STT
+     * Used by IAudioStorage for format conversion
+     */
+    getSTTInputFormat() {
+        return {
+            format: 'wav',
+            sampleRate: 16000,
+            channels: 1,
+            bitDepth: 16
+        };
+    }
+}
+exports.GeminiVoiceAdapter = GeminiVoiceAdapter;
+//# sourceMappingURL=GeminiVoiceAdapter.js.map

package/dist/adapters/voice/GeminiVoiceAdapter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"GeminiVoiceAdapter.js","sourceRoot":"","sources":["../../../src/adapters/voice/GeminiVoiceAdapter.ts"],"names":[],"mappings":";;;AAEA;;;;GAIG;AACH,MAAa,kBAAkB;IAI3B,YAAY,MAAc,EAAE,WAAyB;QACjD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,WAAW,GAAG,WAAW,CAAC;IACnC,CAAC;IAED;;;;;OAKG;IACH,KAAK,CAAC,YAAY,CAAC,WAAmB,EAAE,MAA2B;QAC/D,MAAM,SAAS,GAAG,MAAM,IAAI,IAAI,CAAC,WAAW,EAAE,GAAG,CAAC;QAClD,MAAM,QAAQ,GAAG,SAAS,EAAE,QAAQ,IAAI,OAAO,CAAC;QAEhD,IAAI,CAAC;YACD,MAAM,QAAQ,GAAG,MAAM,KAAK,CACxB,qGAAqG,IAAI,CAAC,MAAM,EAAE,EAClH;gBACI,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE;gBAC/C,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;oBACjB,QAAQ,EAAE,CAAC;4BACP,IAAI,EAAE,MAAM;4BACZ,KAAK,EAAE;gCACH,EAAE,IAAI,EAAE,qGAAqG,EAAE;gCAC/G;oCACI,UAAU,EAAE;wCACR,QAAQ,EAAE,WAAW;wCACrB,IAAI,EAAE,WAAW;qCACpB;iCACJ;6BACJ;yBACJ,CAAC;oBACF,gBAAgB,EAAE;wBACd,WAAW,EAAE,GAAG,EAAE,+CAA+C;qBACpE;iBACJ,CAAC;aACL,CACJ,CAAC;YAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;gBACf,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;gBACxC,MAAM,IAAI,KAAK,CAAC,yBAAyB,QAAQ,CAAC,MAAM,MAAM,SAAS,EAAE,CAAC,CAAC;YAC/E,CAAC;YAED,MAAM,IAAI,GAAQ,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACxC,MAAM,UAAU,GAAG,IAAI,CAAC,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC;YAEzE,IAAI,CAAC,UAAU,EAAE,CAAC;gBACd,MAAM,IAAI,KAAK,CAAC,wCAAwC,CAAC,CAAC;YAC9D,CAAC;YAED,OAAO,UAAU,CAAC;QACtB,CAAC;QAAC,OAAO,KAAU,EAAE,CAAC;YAClB,OAAO,CAAC,KAAK,CAAC,uBAAuB,EAAE,KAAK,CAAC,CAAC;YAC9C,MAAM,IAAI,KAAK,CAAC,eAAe,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;QACpD,CAAC;IACL,CAAC;IAED;;;;;OAKG;IACH,KAAK,CAAC,YAAY,CAAC,IAAY,EAAE,MAA2B;QACxD,MAAM,SAAS,GAAG,MAAM,IAAI,IAAI,CAAC,WAAW,EAAE,GAAG,CAAC;QAClD,MAAM,KAAK,GAAG,SAAS,EAAE,KAAK,IAAI,MAAM,CAAC,CAAC,yBAAyB;QAEnE,IAAI,CAAC;YACD,kDAAkD;YAClD,MAAM,QAAQ,GAAG,MAAM,KAAK,CACxB,qGAAqG,IAAI,CAAC,MAAM,EAAE,EAClH;gBACI,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE;gBAC/C,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;oBACjB,QAAQ,EAAE,CAAC;4BACP,IAAI,EAAE,MAAM;4BACZ,KAAK,EAAE,CAAC,EAAE,IAAI,EAAE,CAAC;yBACpB,CAAC;oBACF,gBAAgB,EAAE;wBACd,kBAAkB,EAAE,CAAC,OAAO,CAAC;wBAC7B,YAAY,EAAE;4BACV,WAAW,EAAE;gCACT,mBAAmB,EAAE;oCACjB,SAAS,EAAE,KAAK;iCACnB;6BACJ;yBACJ;qBACJ;iBACJ,CAAC;aACL,CACJ,CAAC;YAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;gBACf,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;gBACxC,MAAM,IAAI,KAAK,CAAC,yBAAyB,QAAQ,CAAC,MAAM,MAAM,SAAS,EAAE,CAAC,CAAC;YAC/E,CAAC;YAED,MAAM,IAAI,GAAQ,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACxC,MAAM,WAAW,GAAG,IAAI,CAAC,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,UAAU,EAAE,IAAI,CAAC;YAEhF,IAAI,CAAC,WAAW,EAAE,CAAC;gBACf,MAAM,IAAI,KAAK,CAAC,wCAAwC,CAAC,CAAC;YAC9D,CAAC;YAED,OAAO,UAAU,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,WAAW,EAAE,QAAQ,CAAC,CAAC,CAAC;QAC/D,CAAC;QAAC,OAAO,KAAU,EAAE,CAAC;YAClB,OAAO,CAAC,KAAK,CAAC,uBAAuB,EAAE,KAAK,CAAC,CAAC;YAC9C,MAAM,IAAI,KAAK,CAAC,eAAe,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;QACpD,CAAC;IACL,CAAC;IAED;;;OAGG;IACH,kBAAkB;QACd,OAAO;YACH,MAAM,EAAE,KAAK;YACb,UAAU,EAAE,KAAK;YACjB,QAAQ,EAAE,CAAC;YACX,QAAQ,EAAE,EAAE;SACf,CAAC;IACN,CAAC;IAED;;;OAGG;IACH,iBAAiB;QACb,OAAO;YACH,MAAM,EAAE,KAAK;YACb,UAAU,EAAE,KAAK;YACjB,QAAQ,EAAE,CAAC;YACX,QAAQ,EAAE,EAAE;SACf,CAAC;IACN,CAAC;CACJ;AAhJD,gDAgJC"}

package/dist/adapters/voice/GoogleTTSAdapter.d.ts

@@ -0,0 +1,61 @@
+import { IVoiceAdapter, VoiceConfig } from '../../interfaces';
+/**
+ * Google Cloud Text-to-Speech Adapter
+ *
+ * Production-ready TTS using Google Cloud Text-to-Speech API
+ * - Supports 70+ languages with 400+ voices
+ * - Male/female voice options
+ * - WaveNet & Neural2 premium voices
+ * - SSML support for advanced control
+ * - Perfect for Persian, Arabic, Asian languages
+ *
+ * Pricing:
+ * - Standard voices: $4 per 1M characters
+ * - WaveNet/Neural2: $16 per 1M characters
+ */
+export declare class GoogleTTSAdapter implements IVoiceAdapter {
+    private apiKey;
+    private voiceConfig?;
+    constructor(apiKey: string, voiceConfig?: VoiceConfig);
+    /**
+     * Speech-to-text (not implemented - use GeminiVoiceAdapter)
+     */
+    speechToText(audioBase64: string, config?: VoiceConfig['stt']): Promise<string>;
+    /**
+     * Text-to-speech using Google Cloud TTS
+     *
+     * Features:
+     * - Multiple audio formats (LINEAR16, MP3, OGG)
+     * - Pitch and speaking rate control
+     * - SSML support
+     * - Volume gain control
+     *
+     * @param text Text to convert to speech (or SSML)
+     * @param config Optional TTS configuration
+     * @returns Audio data as Uint8Array
+     */
+    textToSpeech(text: string, config?: VoiceConfig['tts']): Promise<Uint8Array>;
+    /**
+     * Extract language code from voice ID
+     * Examples: 'fa-IR-Standard-A' → 'fa-IR', 'ar-XA-Wavenet-B' → 'ar-XA'
+     */
+    private extractLanguageCode;
+    /**
+     * Validate voice ID format (basic check)
+     */
+    private isValidVoiceId;
+    /**
+     * Check if text is SSML (Speech Synthesis Markup Language)
+     */
+    private isSSML;
+    /**
+     * Extract gender hint from voice variant (A-D typically female, E+ typically male)
+     * This is a heuristic and may not always be accurate
+     */
+    private extractGender;
+    /**
+     * Clamp value between min and max
+     */
+    private clamp;
+}
+//# sourceMappingURL=GoogleTTSAdapter.d.ts.map

package/dist/adapters/voice/GoogleTTSAdapter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"GoogleTTSAdapter.d.ts","sourceRoot":"","sources":["../../../src/adapters/voice/GoogleTTSAdapter.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,aAAa,EAAE,WAAW,EAAE,MAAM,kBAAkB,CAAC;AAE9D;;;;;;;;;;;;;GAaG;AACH,qBAAa,gBAAiB,YAAW,aAAa;IAClD,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,WAAW,CAAC,CAAc;gBAEtB,MAAM,EAAE,MAAM,EAAE,WAAW,CAAC,EAAE,WAAW;IAKrD;;OAEG;IACG,YAAY,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,CAAC,KAAK,CAAC,GAAG,OAAO,CAAC,MAAM,CAAC;IAIrF;;;;;;;;;;;;OAYG;IACG,YAAY,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,CAAC,KAAK,CAAC,GAAG,OAAO,CAAC,UAAU,CAAC;IAiFlF;;;OAGG;IACH,OAAO,CAAC,mBAAmB;IAS3B;;OAEG;IACH,OAAO,CAAC,cAAc;IAOtB;;OAEG;IACH,OAAO,CAAC,MAAM;IAId;;;OAGG;IACH,OAAO,CAAC,aAAa;IAgBrB;;OAEG;IACH,OAAO,CAAC,KAAK;CAGhB"}

package/dist/adapters/voice/GoogleTTSAdapter.js

@@ -0,0 +1,165 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.GoogleTTSAdapter = void 0;
+/**
+ * Google Cloud Text-to-Speech Adapter
+ *
+ * Production-ready TTS using Google Cloud Text-to-Speech API
+ * - Supports 70+ languages with 400+ voices
+ * - Male/female voice options
+ * - WaveNet & Neural2 premium voices
+ * - SSML support for advanced control
+ * - Perfect for Persian, Arabic, Asian languages
+ *
+ * Pricing:
+ * - Standard voices: $4 per 1M characters
+ * - WaveNet/Neural2: $16 per 1M characters
+ */
+class GoogleTTSAdapter {
+    constructor(apiKey, voiceConfig) {
+        this.apiKey = apiKey;
+        this.voiceConfig = voiceConfig;
+    }
+    /**
+     * Speech-to-text (not implemented - use GeminiVoiceAdapter)
+     */
+    async speechToText(audioBase64, config) {
+        throw new Error('GoogleTTSAdapter does not support STT. Use GeminiVoiceAdapter for STT.');
+    }
+    /**
+     * Text-to-speech using Google Cloud TTS
+     *
+     * Features:
+     * - Multiple audio formats (LINEAR16, MP3, OGG)
+     * - Pitch and speaking rate control
+     * - SSML support
+     * - Volume gain control
+     *
+     * @param text Text to convert to speech (or SSML)
+     * @param config Optional TTS configuration
+     * @returns Audio data as Uint8Array
+     */
+    async textToSpeech(text, config) {
+        const ttsConfig = config || this.voiceConfig?.tts;
+        const voiceId = ttsConfig?.voice || 'en-US-Standard-A';
+        const languageCode = this.extractLanguageCode(voiceId);
+        const startTime = Date.now();
+        try {
+            // Validate text
+            if (!text || text.trim().length === 0) {
+                throw new Error('Text input is empty');
+            }
+            // Validate voice ID format
+            if (!this.isValidVoiceId(voiceId)) {
+                console.warn(`[GoogleTTSAdapter] Invalid voice ID format: ${voiceId}. Using default.`);
+            }
+            // Build request payload
+            const payload = {
+                input: this.isSSML(text) ? { ssml: text } : { text },
+                voice: {
+                    languageCode: languageCode,
+                    name: voiceId,
+                    // Optional: Add SSML gender hint for better selection
+                    ...(this.extractGender(voiceId) && { ssmlGender: this.extractGender(voiceId) })
+                },
+                audioConfig: {
+                    audioEncoding: 'LINEAR16', // PCM 16-bit (compatible with most systems)
+                    speakingRate: this.clamp(ttsConfig?.speakingRate || 1.0, 0.25, 4.0),
+                    pitch: this.clamp(ttsConfig?.pitch || 0.0, -20.0, 20.0),
+                    volumeGainDb: 0.0, // Default volume
+                    sampleRateHertz: 24000 // High quality
+                }
+            };
+            const response = await fetch(`https://texttospeech.googleapis.com/v1/text:synthesize?key=${this.apiKey}`, {
+                method: 'POST',
+                headers: { 'Content-Type': 'application/json' },
+                body: JSON.stringify(payload)
+            });
+            if (!response.ok) {
+                const errorData = await response.json().catch(() => ({ error: { message: 'Unknown error' } }));
+                const errorMessage = errorData.error?.message || `HTTP ${response.status}`;
+                // Provide helpful error messages
+                if (response.status === 400) {
+                    throw new Error(`Invalid request: ${errorMessage}. Check voice ID and language code.`);
+                }
+                else if (response.status === 403) {
+                    throw new Error(`API not enabled or invalid key: ${errorMessage}. Enable Cloud Text-to-Speech API.`);
+                }
+                else if (response.status === 429) {
+                    throw new Error(`Quota exceeded: ${errorMessage}. Check your Google Cloud quota.`);
+                }
+                else {
+                    throw new Error(`Google TTS API Error (${response.status}): ${errorMessage}`);
+                }
+            }
+            const json = await response.json();
+            const audioBase64 = json.audioContent;
+            if (!audioBase64) {
+                throw new Error('No audio data received from Google TTS');
+            }
+            const audioData = Uint8Array.from(Buffer.from(audioBase64, 'base64'));
+            const duration = Date.now() - startTime;
+            console.log(`[GoogleTTSAdapter] TTS successful: ${audioData.length} bytes, ${text.length} chars, ${duration}ms (voice: ${voiceId})`);
+            return audioData;
+        }
+        catch (error) {
+            const duration = Date.now() - startTime;
+            console.error(`[GoogleTTSAdapter] TTS Error after ${duration}ms:`, error.message);
+            throw new Error(`Google TTS failed: ${error.message}`);
+        }
+    }
+    /**
+     * Extract language code from voice ID
+     * Examples: 'fa-IR-Standard-A' → 'fa-IR', 'ar-XA-Wavenet-B' → 'ar-XA'
+     */
+    extractLanguageCode(voiceId) {
+        const parts = voiceId.split('-');
+        if (parts.length >= 2) {
+            return `${parts[0]}-${parts[1]}`;
+        }
+        console.warn(`[GoogleTTSAdapter] Could not extract language code from: ${voiceId}`);
+        return 'en-US'; // Fallback
+    }
+    /**
+     * Validate voice ID format (basic check)
+     */
+    isValidVoiceId(voiceId) {
+        // Google voice IDs format: {lang}-{region}-{type}-{variant}
+        // e.g., 'en-US-Standard-A', 'fa-IR-Wavenet-B'
+        const pattern = /^[a-z]{2,3}-[A-Z]{2}-(Standard|Wavenet|Neural2|Studio|Journey|News|Polyglot)-[A-Za-z]$/;
+        return pattern.test(voiceId);
+    }
+    /**
+     * Check if text is SSML (Speech Synthesis Markup Language)
+     */
+    isSSML(text) {
+        return text.trim().startsWith('<speak>') && text.trim().endsWith('</speak>');
+    }
+    /**
+     * Extract gender hint from voice variant (A-D typically female, E+ typically male)
+     * This is a heuristic and may not always be accurate
+     */
+    extractGender(voiceId) {
+        const variant = voiceId.split('-').pop();
+        if (!variant)
+            return null;
+        // Google's general pattern (not guaranteed):
+        // A, B, C, D = Female
+        // E, F, G, H = Male
+        const femaleVariants = ['A', 'B', 'C', 'D'];
+        const maleVariants = ['E', 'F', 'G', 'H'];
+        if (femaleVariants.includes(variant))
+            return 'FEMALE';
+        if (maleVariants.includes(variant))
+            return 'MALE';
+        return null;
+    }
+    /**
+     * Clamp value between min and max
+     */
+    clamp(value, min, max) {
+        return Math.max(min, Math.min(max, value));
+    }
+}
+exports.GoogleTTSAdapter = GoogleTTSAdapter;
+//# sourceMappingURL=GoogleTTSAdapter.js.map

package/dist/adapters/voice/GoogleTTSAdapter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"GoogleTTSAdapter.js","sourceRoot":"","sources":["../../../src/adapters/voice/GoogleTTSAdapter.ts"],"names":[],"mappings":";;;AAEA;;;;;;;;;;;;;GAaG;AACH,MAAa,gBAAgB;IAIzB,YAAY,MAAc,EAAE,WAAyB;QACjD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,WAAW,GAAG,WAAW,CAAC;IACnC,CAAC;IAED;;OAEG;IACH,KAAK,CAAC,YAAY,CAAC,WAAmB,EAAE,MAA2B;QAC/D,MAAM,IAAI,KAAK,CAAC,wEAAwE,CAAC,CAAC;IAC9F,CAAC;IAED;;;;;;;;;;;;OAYG;IACH,KAAK,CAAC,YAAY,CAAC,IAAY,EAAE,MAA2B;QACxD,MAAM,SAAS,GAAG,MAAM,IAAI,IAAI,CAAC,WAAW,EAAE,GAAG,CAAC;QAClD,MAAM,OAAO,GAAG,SAAS,EAAE,KAAK,IAAI,kBAAkB,CAAC;QACvD,MAAM,YAAY,GAAG,IAAI,CAAC,mBAAmB,CAAC,OAAO,CAAC,CAAC;QAEvD,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAE7B,IAAI,CAAC;YACD,gBAAgB;YAChB,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;gBACpC,MAAM,IAAI,KAAK,CAAC,qBAAqB,CAAC,CAAC;YAC3C,CAAC;YAED,2BAA2B;YAC3B,IAAI,CAAC,IAAI,CAAC,cAAc,CAAC,OAAO,CAAC,EAAE,CAAC;gBAChC,OAAO,CAAC,IAAI,CAAC,+CAA+C,OAAO,kBAAkB,CAAC,CAAC;YAC3F,CAAC;YAED,wBAAwB;YACxB,MAAM,OAAO,GAAG;gBACZ,KAAK,EAAE,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE;gBACpD,KAAK,EAAE;oBACH,YAAY,EAAE,YAAY;oBAC1B,IAAI,EAAE,OAAO;oBACb,sDAAsD;oBACtD,GAAG,CAAC,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,IAAI,EAAE,UAAU,EAAE,IAAI,CAAC,aAAa,CAAC,OAAO,CAAC,EAAE,CAAC;iBAClF;gBACD,WAAW,EAAE;oBACT,aAAa,EAAE,UAAU,EAAE,4CAA4C;oBACvE,YAAY,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,YAAY,IAAI,GAAG,EAAE,IAAI,EAAE,GAAG,CAAC;oBACnE,KAAK,EAAE,IAAI,CAAC,KAAK,CAAC,SAAS,EAAE,KAAK,IAAI,GAAG,EAAE,CAAC,IAAI,EAAE,IAAI,CAAC;oBACvD,YAAY,EAAE,GAAG,EAAE,iBAAiB;oBACpC,eAAe,EAAE,KAAK,CAAC,eAAe;iBACzC;aACJ,CAAC;YAEF,MAAM,QAAQ,GAAG,MAAM,KAAK,CACxB,8DAA8D,IAAI,CAAC,MAAM,EAAE,EAC3E;gBACI,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE;gBAC/C,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,OAAO,CAAC;aAChC,CACJ,CAAC;YAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;gBACf,MAAM,SAAS,GAAQ,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC,CAAC,EAAE,KAAK,EAAE,EAAE,OAAO,EAAE,eAAe,EAAE,EAAE,CAAC,CAAC,CAAC;gBACpG,MAAM,YAAY,GAAG,SAAS,CAAC,KAAK,EAAE,OAAO,IAAI,QAAQ,QAAQ,CAAC,MAAM,EAAE,CAAC;gBAE3E,iCAAiC;gBACjC,IAAI,QAAQ,CAAC,MAAM,KAAK,GAAG,EAAE,CAAC;oBAC1B,MAAM,IAAI,KAAK,CAAC,oBAAoB,YAAY,qCAAqC,CAAC,CAAC;gBAC3F,CAAC;qBAAM,IAAI,QAAQ,CAAC,MAAM,KAAK,GAAG,EAAE,CAAC;oBACjC,MAAM,IAAI,KAAK,CAAC,mCAAmC,YAAY,oCAAoC,CAAC,CAAC;gBACzG,CAAC;qBAAM,IAAI,QAAQ,CAAC,MAAM,KAAK,GAAG,EAAE,CAAC;oBACjC,MAAM,IAAI,KAAK,CAAC,mBAAmB,YAAY,kCAAkC,CAAC,CAAC;gBACvF,CAAC;qBAAM,CAAC;oBACJ,MAAM,IAAI,KAAK,CAAC,yBAAyB,QAAQ,CAAC,MAAM,MAAM,YAAY,EAAE,CAAC,CAAC;gBAClF,CAAC;YACL,CAAC;YAED,MAAM,IAAI,GAAQ,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACxC,MAAM,WAAW,GAAG,IAAI,CAAC,YAAY,CAAC;YAEtC,IAAI,CAAC,WAAW,EAAE,CAAC;gBACf,MAAM,IAAI,KAAK,CAAC,wCAAwC,CAAC,CAAC;YAC9D,CAAC;YAED,MAAM,SAAS,GAAG,UAAU,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,WAAW,EAAE,QAAQ,CAAC,CAAC,CAAC;YACtE,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS,CAAC;YAExC,OAAO,CAAC,GAAG,CAAC,sCAAsC,SAAS,CAAC,MAAM,WAAW,IAAI,CAAC,MAAM,WAAW,QAAQ,cAAc,OAAO,GAAG,CAAC,CAAC;YAErI,OAAO,SAAS,CAAC;QACrB,CAAC;QAAC,OAAO,KAAU,EAAE,CAAC;YAClB,MAAM,QAAQ,GAAG,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS,CAAC;YACxC,OAAO,CAAC,KAAK,CAAC,sCAAsC,QAAQ,KAAK,EAAE,KAAK,CAAC,OAAO,CAAC,CAAC;YAClF,MAAM,IAAI,KAAK,CAAC,sBAAsB,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;QAC3D,CAAC;IACL,CAAC;IAED;;;OAGG;IACK,mBAAmB,CAAC,OAAe;QACvC,MAAM,KAAK,GAAG,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC;QACjC,IAAI,KAAK,CAAC,MAAM,IAAI,CAAC,EAAE,CAAC;YACpB,OAAO,GAAG,KAAK,CAAC,CAAC,CAAC,IAAI,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC;QACrC,CAAC;QACD,OAAO,CAAC,IAAI,CAAC,4DAA4D,OAAO,EAAE,CAAC,CAAC;QACpF,OAAO,OAAO,CAAC,CAAC,WAAW;IAC/B,CAAC;IAED;;OAEG;IACK,cAAc,CAAC,OAAe;QAClC,4DAA4D;QAC5D,8CAA8C;QAC9C,MAAM,OAAO,GAAG,wFAAwF,CAAC;QACzG,OAAO,OAAO,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC;IACjC,CAAC;IAED;;OAEG;IACK,MAAM,CAAC,IAAY;QACvB,OAAO,IAAI,CAAC,IAAI,EAAE,CAAC,UAAU,CAAC,SAAS,CAAC,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC,QAAQ,CAAC,UAAU,CAAC,CAAC;IACjF,CAAC;IAED;;;OAGG;IACK,aAAa,CAAC,OAAe;QACjC,MAAM,OAAO,GAAG,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,GAAG,EAAE,CAAC;QACzC,IAAI,CAAC,OAAO;YAAE,OAAO,IAAI,CAAC;QAE1B,6CAA6C;QAC7C,sBAAsB;QACtB,oBAAoB;QACpB,MAAM,cAAc,GAAG,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;QAC5C,MAAM,YAAY,GAAG,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,GAAG,CAAC,CAAC;QAE1C,IAAI,cAAc,CAAC,QAAQ,CAAC,OAAO,CAAC;YAAE,OAAO,QAAQ,CAAC;QACtD,IAAI,YAAY,CAAC,QAAQ,CAAC,OAAO,CAAC;YAAE,OAAO,MAAM,CAAC;QAElD,OAAO,IAAI,CAAC;IAChB,CAAC;IAED;;OAEG;IACK,KAAK,CAAC,KAAa,EAAE,GAAW,EAAE,GAAW;QACjD,OAAO,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,KAAK,CAAC,CAAC,CAAC;IAC/C,CAAC;CACJ;AAtKD,4CAsKC"}

package/dist/adapters/voice/GoogleVoiceAdapter.d.ts

@@ -0,0 +1,33 @@
+import { IVoiceAdapter, VoiceConfig, AudioFormatMetadata } from '../../interfaces';
+/**
+ * Google Voice Adapter
+ * Uses Google services exclusively:
+ * - Gemini API for STT (Speech-to-Text)
+ * - Google Cloud Text-to-Speech API for TTS
+ *
+ * Outputs standard LINEAR16 PCM audio for perfect concatenation
+ */
+export declare class GoogleVoiceAdapter implements IVoiceAdapter {
+    private apiKey;
+    private voiceConfig?;
+    private ttsClient;
+    constructor(apiKey: string, voiceConfig?: VoiceConfig);
+    /**
+     * Convert speech (audio) to text using Gemini's audio input capability
+     */
+    speechToText(audioBase64: string, config?: VoiceConfig['stt']): Promise<string>;
+    /**
+     * Convert text to speech using Google Cloud Text-to-Speech API
+     * Supports both API key and service account authentication
+     */
+    textToSpeech(text: string, config?: VoiceConfig['tts']): Promise<Uint8Array>;
+    /**
+     * Declare the audio format produced by Google TTS
+     */
+    getTTSOutputFormat(): AudioFormatMetadata;
+    /**
+     * Declare the expected audio format for Gemini STT
+     */
+    getSTTInputFormat(): AudioFormatMetadata;
+}
+//# sourceMappingURL=GoogleVoiceAdapter.d.ts.map

package/dist/adapters/voice/GoogleVoiceAdapter.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"GoogleVoiceAdapter.d.ts","sourceRoot":"","sources":["../../../src/adapters/voice/GoogleVoiceAdapter.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,aAAa,EAAE,WAAW,EAAE,mBAAmB,EAAE,MAAM,kBAAkB,CAAC;AAGnF;;;;;;;GAOG;AACH,qBAAa,kBAAmB,YAAW,aAAa;IACpD,OAAO,CAAC,MAAM,CAAS;IACvB,OAAO,CAAC,WAAW,CAAC,CAAc;IAClC,OAAO,CAAC,SAAS,CAAqB;gBAE1B,MAAM,EAAE,MAAM,EAAE,WAAW,CAAC,EAAE,WAAW;IAMrD;;OAEG;IACG,YAAY,CAAC,WAAW,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,CAAC,KAAK,CAAC,GAAG,OAAO,CAAC,MAAM,CAAC;IAiDrF;;;OAGG;IACG,YAAY,CAAC,IAAI,EAAE,MAAM,EAAE,MAAM,CAAC,EAAE,WAAW,CAAC,KAAK,CAAC,GAAG,OAAO,CAAC,UAAU,CAAC;IA0ElF;;OAEG;IACH,kBAAkB,IAAI,mBAAmB;IASzC;;OAEG;IACH,iBAAiB,IAAI,mBAAmB;CAQ3C"}

package/dist/adapters/voice/GoogleVoiceAdapter.js

@@ -0,0 +1,155 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.GoogleVoiceAdapter = void 0;
+const text_to_speech_1 = require("@google-cloud/text-to-speech");
+/**
+ * Google Voice Adapter
+ * Uses Google services exclusively:
+ * - Gemini API for STT (Speech-to-Text)
+ * - Google Cloud Text-to-Speech API for TTS
+ *
+ * Outputs standard LINEAR16 PCM audio for perfect concatenation
+ */
+class GoogleVoiceAdapter {
+    constructor(apiKey, voiceConfig) {
+        this.apiKey = apiKey;
+        this.voiceConfig = voiceConfig;
+        this.ttsClient = new text_to_speech_1.TextToSpeechClient();
+    }
+    /**
+     * Convert speech (audio) to text using Gemini's audio input capability
+     */
+    async speechToText(audioBase64, config) {
+        const sttConfig = config || this.voiceConfig?.stt;
+        const language = sttConfig?.language || 'en-US';
+        try {
+            const response = await fetch(`https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-lite:generateContent?key=${this.apiKey}`, {
+                method: 'POST',
+                headers: { 'Content-Type': 'application/json' },
+                body: JSON.stringify({
+                    contents: [{
+                            role: 'user',
+                            parts: [
+                                { text: "Transcribe the following audio exactly. Do not respond to the audio, just output the transcription." },
+                                {
+                                    inlineData: {
+                                        mimeType: 'audio/wav',
+                                        data: audioBase64
+                                    }
+                                }
+                            ]
+                        }],
+                    generationConfig: {
+                        temperature: 0.1,
+                    }
+                })
+            });
+            if (!response.ok) {
+                const errorText = await response.text();
+                throw new Error(`Gemini STT API Error: ${response.status} - ${errorText}`);
+            }
+            const json = await response.json();
+            const transcript = json.candidates?.[0]?.content?.parts?.[0]?.text || '';
+            if (!transcript) {
+                throw new Error('No transcript returned from Gemini STT');
+            }
+            return transcript;
+        }
+        catch (error) {
+            console.error('Speech-to-Text Error:', error);
+            throw new Error(`STT failed: ${error.message}`);
+        }
+    }
+    /**
+     * Convert text to speech using Google Cloud Text-to-Speech API
+     * Supports both API key and service account authentication
+     */
+    async textToSpeech(text, config) {
+        const ttsConfig = config || this.voiceConfig?.tts;
+        const languageCode = ttsConfig?.language || 'en-US';
+        const voiceName = ttsConfig?.voice || 'en-US-Chirp3-HD-Leda'; // Default female voice
+        try {
+            // Try API key authentication first (simpler)
+            const googleCloudApiKey = process.env.GOOGLE_CLOUD_API_KEY || process.env.GOOGLE_AI_API_KEY;
+            if (googleCloudApiKey) {
+                // Use REST API with API key
+                const response = await fetch(`https://texttospeech.googleapis.com/v1/text:synthesize?key=${googleCloudApiKey}`, {
+                    method: 'POST',
+                    headers: { 'Content-Type': 'application/json' },
+                    body: JSON.stringify({
+                        input: { text },
+                        voice: {
+                            languageCode: languageCode,
+                            name: voiceName
+                        },
+                        audioConfig: {
+                            audioEncoding: 'LINEAR16',
+                            sampleRateHertz: 16000,
+                            pitch: 0,
+                            speakingRate: 1.2
+                        }
+                    })
+                });
+                if (!response.ok) {
+                    const errorText = await response.text();
+                    throw new Error(`Google TTS API Error: ${response.status} - ${errorText}`);
+                }
+                const json = await response.json();
+                const audioContent = json.audioContent;
+                if (!audioContent) {
+                    throw new Error('No audio content received from Google TTS');
+                }
+                // audioContent is base64-encoded
+                return new Uint8Array(Buffer.from(audioContent, 'base64'));
+            }
+            else {
+                // Fall back to service account authentication
+                const [response] = await this.ttsClient.synthesizeSpeech({
+                    input: { text },
+                    voice: {
+                        languageCode: languageCode,
+                        name: voiceName
+                    },
+                    audioConfig: {
+                        audioEncoding: 'LINEAR16',
+                        sampleRateHertz: 16000,
+                        pitch: 0,
+                        speakingRate: 1.2
+                    }
+                });
+                if (!response.audioContent) {
+                    throw new Error('No audio content received from Google TTS');
+                }
+                return new Uint8Array(response.audioContent);
+            }
+        }
+        catch (error) {
+            console.error('Text-to-Speech Error:', error);
+            throw new Error(`TTS failed: ${error.message}`);
+        }
+    }
+    /**
+     * Declare the audio format produced by Google TTS
+     */
+    getTTSOutputFormat() {
+        return {
+            format: 'wav',
+            sampleRate: 16000,
+            channels: 1,
+            bitDepth: 16
+        };
+    }
+    /**
+     * Declare the expected audio format for Gemini STT
+     */
+    getSTTInputFormat() {
+        return {
+            format: 'wav',
+            sampleRate: 16000,
+            channels: 1,
+            bitDepth: 16
+        };
+    }
+}
+exports.GoogleVoiceAdapter = GoogleVoiceAdapter;
+//# sourceMappingURL=GoogleVoiceAdapter.js.map

package/dist/adapters/voice/GoogleVoiceAdapter.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"GoogleVoiceAdapter.js","sourceRoot":"","sources":["../../../src/adapters/voice/GoogleVoiceAdapter.ts"],"names":[],"mappings":";;;AACA,iEAAkE;AAElE;;;;;;;GAOG;AACH,MAAa,kBAAkB;IAK3B,YAAY,MAAc,EAAE,WAAyB;QACjD,IAAI,CAAC,MAAM,GAAG,MAAM,CAAC;QACrB,IAAI,CAAC,WAAW,GAAG,WAAW,CAAC;QAC/B,IAAI,CAAC,SAAS,GAAG,IAAI,mCAAkB,EAAE,CAAC;IAC9C,CAAC;IAED;;OAEG;IACH,KAAK,CAAC,YAAY,CAAC,WAAmB,EAAE,MAA2B;QAC/D,MAAM,SAAS,GAAG,MAAM,IAAI,IAAI,CAAC,WAAW,EAAE,GAAG,CAAC;QAClD,MAAM,QAAQ,GAAG,SAAS,EAAE,QAAQ,IAAI,OAAO,CAAC;QAEhD,IAAI,CAAC;YACD,MAAM,QAAQ,GAAG,MAAM,KAAK,CACxB,qGAAqG,IAAI,CAAC,MAAM,EAAE,EAClH;gBACI,MAAM,EAAE,MAAM;gBACd,OAAO,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE;gBAC/C,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;oBACjB,QAAQ,EAAE,CAAC;4BACP,IAAI,EAAE,MAAM;4BACZ,KAAK,EAAE;gCACH,EAAE,IAAI,EAAE,qGAAqG,EAAE;gCAC/G;oCACI,UAAU,EAAE;wCACR,QAAQ,EAAE,WAAW;wCACrB,IAAI,EAAE,WAAW;qCACpB;iCACJ;6BACJ;yBACJ,CAAC;oBACF,gBAAgB,EAAE;wBACd,WAAW,EAAE,GAAG;qBACnB;iBACJ,CAAC;aACL,CACJ,CAAC;YAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;gBACf,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;gBACxC,MAAM,IAAI,KAAK,CAAC,yBAAyB,QAAQ,CAAC,MAAM,MAAM,SAAS,EAAE,CAAC,CAAC;YAC/E,CAAC;YAED,MAAM,IAAI,GAAQ,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;YACxC,MAAM,UAAU,GAAG,IAAI,CAAC,UAAU,EAAE,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,KAAK,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,IAAI,EAAE,CAAC;YAEzE,IAAI,CAAC,UAAU,EAAE,CAAC;gBACd,MAAM,IAAI,KAAK,CAAC,wCAAwC,CAAC,CAAC;YAC9D,CAAC;YAED,OAAO,UAAU,CAAC;QACtB,CAAC;QAAC,OAAO,KAAU,EAAE,CAAC;YAClB,OAAO,CAAC,KAAK,CAAC,uBAAuB,EAAE,KAAK,CAAC,CAAC;YAC9C,MAAM,IAAI,KAAK,CAAC,eAAe,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;QACpD,CAAC;IACL,CAAC;IAED;;;OAGG;IACH,KAAK,CAAC,YAAY,CAAC,IAAY,EAAE,MAA2B;QACxD,MAAM,SAAS,GAAG,MAAM,IAAI,IAAI,CAAC,WAAW,EAAE,GAAG,CAAC;QAClD,MAAM,YAAY,GAAG,SAAS,EAAE,QAAQ,IAAI,OAAO,CAAC;QACpD,MAAM,SAAS,GAAG,SAAS,EAAE,KAAK,IAAI,sBAAsB,CAAC,CAAC,uBAAuB;QAErF,IAAI,CAAC;YACD,6CAA6C;YAC7C,MAAM,iBAAiB,GAAG,OAAO,CAAC,GAAG,CAAC,oBAAoB,IAAI,OAAO,CAAC,GAAG,CAAC,iBAAiB,CAAC;YAE5F,IAAI,iBAAiB,EAAE,CAAC;gBACpB,4BAA4B;gBAC5B,MAAM,QAAQ,GAAG,MAAM,KAAK,CACxB,8DAA8D,iBAAiB,EAAE,EACjF;oBACI,MAAM,EAAE,MAAM;oBACd,OAAO,EAAE,EAAE,cAAc,EAAE,kBAAkB,EAAE;oBAC/C,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC;wBACjB,KAAK,EAAE,EAAE,IAAI,EAAE;wBACf,KAAK,EAAE;4BACH,YAAY,EAAE,YAAY;4BAC1B,IAAI,EAAE,SAAS;yBAClB;wBACD,WAAW,EAAE;4BACT,aAAa,EAAE,UAAU;4BACzB,eAAe,EAAE,KAAK;4BACtB,KAAK,EAAE,CAAC;4BACR,YAAY,EAAE,GAAG;yBACpB;qBACJ,CAAC;iBACL,CACJ,CAAC;gBAEF,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;oBACf,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;oBACxC,MAAM,IAAI,KAAK,CAAC,yBAAyB,QAAQ,CAAC,MAAM,MAAM,SAAS,EAAE,CAAC,CAAC;gBAC/E,CAAC;gBAED,MAAM,IAAI,GAAQ,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;gBACxC,MAAM,YAAY,GAAG,IAAI,CAAC,YAAY,CAAC;gBAEvC,IAAI,CAAC,YAAY,EAAE,CAAC;oBAChB,MAAM,IAAI,KAAK,CAAC,2CAA2C,CAAC,CAAC;gBACjE,CAAC;gBAED,iCAAiC;gBACjC,OAAO,IAAI,UAAU,CAAC,MAAM,CAAC,IAAI,CAAC,YAAY,EAAE,QAAQ,CAAC,CAAC,CAAC;YAC/D,CAAC;iBAAM,CAAC;gBACJ,8CAA8C;gBAC9C,MAAM,CAAC,QAAQ,CAAC,GAAG,MAAM,IAAI,CAAC,SAAS,CAAC,gBAAgB,CAAC;oBACrD,KAAK,EAAE,EAAE,IAAI,EAAE;oBACf,KAAK,EAAE;wBACH,YAAY,EAAE,YAAY;wBAC1B,IAAI,EAAE,SAAS;qBAClB;oBACD,WAAW,EAAE;wBACT,aAAa,EAAE,UAAU;wBACzB,eAAe,EAAE,KAAK;wBACtB,KAAK,EAAE,CAAC;wBACR,YAAY,EAAE,GAAG;qBACpB;iBACJ,CAAC,CAAC;gBAEH,IAAI,CAAC,QAAQ,CAAC,YAAY,EAAE,CAAC;oBACzB,MAAM,IAAI,KAAK,CAAC,2CAA2C,CAAC,CAAC;gBACjE,CAAC;gBAED,OAAO,IAAI,UAAU,CAAC,QAAQ,CAAC,YAAsB,CAAC,CAAC;YAC3D,CAAC;QACL,CAAC;QAAC,OAAO,KAAU,EAAE,CAAC;YAClB,OAAO,CAAC,KAAK,CAAC,uBAAuB,EAAE,KAAK,CAAC,CAAC;YAC9C,MAAM,IAAI,KAAK,CAAC,eAAe,KAAK,CAAC,OAAO,EAAE,CAAC,CAAC;QACpD,CAAC;IACL,CAAC;IAED;;OAEG;IACH,kBAAkB;QACd,OAAO;YACH,MAAM,EAAE,KAAK;YACb,UAAU,EAAE,KAAK;YACjB,QAAQ,EAAE,CAAC;YACX,QAAQ,EAAE,EAAE;SACf,CAAC;IACN,CAAC;IAED;;OAEG;IACH,iBAAiB;QACb,OAAO;YACH,MAAM,EAAE,KAAK;YACb,UAAU,EAAE,KAAK;YACjB,QAAQ,EAAE,CAAC;YACX,QAAQ,EAAE,EAAE;SACf,CAAC;IACN,CAAC;CACJ;AApKD,gDAoKC"}