@0xmaxma/claude-gateway 1.8.13 → 2.0.0

package/dist/voice/providers/elevenlabs-stt.js

@@ -0,0 +1,92 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.ElevenLabsStt = void 0;
+const errors_1 = require("../errors");
+const types_1 = require("../types");
+const queue_1 = require("../queue");
+const socket_1 = require("./socket");
+class ElevenLabsStt {
+    constructor(key, model = 'scribe_v2_realtime', connect = socket_1.openProviderSocket, id = 'elevenlabs') {
+        this.key = key;
+        this.model = model;
+        this.connect = connect;
+        this.id = id;
+        this.capabilities = { partials: true, manualCommit: true, speechActivityEvents: false };
+    }
+    async open(options) {
+        (0, types_1.validateAudioFormat)(options.format);
+        if (!this.key)
+            throw new types_1.VoiceError('STT_CREDENTIALS_MISSING');
+        const url = new URL('wss://api.elevenlabs.io/v1/speech-to-text/realtime');
+        url.searchParams.set('model_id', this.model);
+        url.searchParams.set('audio_format', `pcm_${options.format.sampleRate}`);
+        url.searchParams.set('commit_strategy', 'manual');
+        url.searchParams.set('include_timestamps', 'false');
+        if (options.language)
+            url.searchParams.set('language_code', options.language);
+        const socket = await this.connect(url.toString(), { 'xi-api-key': this.key }, options.signal);
+        const events = new queue_1.BoundedQueue(65536, e => JSON.stringify(e).length);
+        let commitId, segment = 0, closed = false;
+        let lastAudioAt = Date.now();
+        const sendAudio = async (frame) => {
+            lastAudioAt = Date.now();
+            await socket.send({ message_type: 'input_audio_chunk', audio_base_64: Buffer.from(frame).toString('base64'), sample_rate: options.format.sampleRate });
+        };
+        // The browser gates uplink during silence. A small silent audio chunk keeps
+        // the recognizer alive while the worker runs; it never commits an input.
+        const keepAlive = setInterval(() => {
+            if (closed || options.signal.aborted || commitId || Date.now() - lastAudioAt < 5000)
+                return;
+            void sendAudio(Buffer.alloc(options.format.sampleRate / 10 * 2)).catch(() => {
+                if (!closed) {
+                    events.close(new types_1.VoiceError('STT_CONNECTION_CLOSED'));
+                    socket.close();
+                }
+            });
+        }, 5000);
+        keepAlive.unref();
+        void (async () => {
+            try {
+                for await (const message of socket.messages) {
+                    if (closed)
+                        break;
+                    if (message.message_type === 'partial_transcript')
+                        events.push({ type: 'partial', segmentId: String(segment), text: String(message.text ?? '') });
+                    else if (message.message_type === 'committed_transcript') {
+                        events.push({ type: 'segment_final', segmentId: String(segment++), text: String(message.text ?? '') });
+                        if (commitId) {
+                            events.push({ type: 'commit_done', commitId });
+                            commitId = undefined;
+                        }
+                    }
+                    else if (message.error)
+                        throw (0, errors_1.providerVoiceError)('STT', undefined, message);
+                }
+                if (!closed)
+                    events.push({ type: 'error', code: 'STT_CONNECTION_CLOSED', retryable: true });
+                events.close();
+            }
+            catch (error) {
+                events.close(error instanceof types_1.VoiceError ? error : new types_1.VoiceError('STT_PROVIDER_ERROR'));
+                socket.close();
+            }
+            finally {
+                closed = true;
+                clearInterval(keepAlive);
+            }
+        })();
+        return {
+            events,
+            pushAudio: sendAudio,
+            commitSegment: async (id) => {
+                if (commitId)
+                    throw new types_1.VoiceError('COMMIT_IN_PROGRESS');
+                commitId = id;
+                await socket.send({ message_type: 'input_audio_chunk', audio_base_64: '', commit: true, sample_rate: options.format.sampleRate });
+            },
+            close: async () => { closed = true; clearInterval(keepAlive); socket.close(); events.close(); },
+        };
+    }
+}
+exports.ElevenLabsStt = ElevenLabsStt;
+//# sourceMappingURL=elevenlabs-stt.js.map

package/dist/voice/providers/elevenlabs-stt.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"elevenlabs-stt.js","sourceRoot":"","sources":["../../../src/voice/providers/elevenlabs-stt.ts"],"names":[],"mappings":";;;AAAA,sCAA+C;AAC/C,oCAA2G;AAC3G,oCAAwC;AACxC,qCAA6D;AAE7D,MAAa,aAAa;IAExB,YAA6B,GAAW,EAAmB,QAAQ,oBAAoB,EAAmB,UAAyB,2BAAkB,EAAW,KAAa,YAAY;QAA5J,QAAG,GAAH,GAAG,CAAQ;QAAmB,UAAK,GAAL,KAAK,CAAuB;QAAmB,YAAO,GAAP,OAAO,CAAoC;QAAW,OAAE,GAAF,EAAE,CAAuB;QADhL,iBAAY,GAAG,EAAE,QAAQ,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,oBAAoB,EAAE,KAAK,EAAE,CAAC;IACgG,CAAC;IAC7L,KAAK,CAAC,IAAI,CAAC,OAAwE;QACjF,IAAA,2BAAmB,EAAC,OAAO,CAAC,MAAM,CAAC,CAAC;QACpC,IAAI,CAAC,IAAI,CAAC,GAAG;YAAE,MAAM,IAAI,kBAAU,CAAC,yBAAyB,CAAC,CAAC;QAC/D,MAAM,GAAG,GAAG,IAAI,GAAG,CAAC,oDAAoD,CAAC,CAAC;QAC1E,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC;QAAC,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,cAAc,EAAE,OAAO,OAAO,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC,CAAC;QACvH,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,iBAAiB,EAAE,QAAQ,CAAC,CAAC;QAAC,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,oBAAoB,EAAE,OAAO,CAAC,CAAC;QACvG,IAAI,OAAO,CAAC,QAAQ;YAAE,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,eAAe,EAAE,OAAO,CAAC,QAAQ,CAAC,CAAC;QAC9E,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,QAAQ,EAAE,EAAE,EAAE,YAAY,EAAE,IAAI,CAAC,GAAG,EAAE,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC;QAC9F,MAAM,MAAM,GAAG,IAAI,oBAAY,CAAW,KAAK,EAAE,CAAC,CAAC,EAAE,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC;QAChF,IAAI,QAA4B,EAAE,OAAO,GAAG,CAAC,EAAE,MAAM,GAAG,KAAK,CAAC;QAC9D,IAAI,WAAW,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAC7B,MAAM,SAAS,GAAG,KAAK,EAAE,KAAiB,EAAE,EAAE;YAC5C,WAAW,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;YACzB,MAAM,MAAM,CAAC,IAAI,CAAC,EAAE,YAAY,EAAE,mBAAmB,EAAE,aAAa,EAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,QAAQ,CAAC,QAAQ,CAAC,EAAE,WAAW,EAAE,OAAO,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC,CAAC;QACzJ,CAAC,CAAC;QACF,4EAA4E;QAC5E,yEAAyE;QACzE,MAAM,SAAS,GAAG,WAAW,CAAC,GAAG,EAAE;YACjC,IAAI,MAAM,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO,IAAI,QAAQ,IAAI,IAAI,CAAC,GAAG,EAAE,GAAG,WAAW,GAAG,IAAI;gBAAE,OAAO;YAC5F,KAAK,SAAS,CAAC,MAAM,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,UAAU,GAAG,EAAE,GAAG,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,GAAG,EAAE;gBAC1E,IAAI,CAAC,MAAM,EAAE,CAAC;oBAAC,MAAM,CAAC,KAAK,CAAC,IAAI,kBAAU,CAAC,uBAAuB,CAAC,CAAC,CAAC;oBAAC,MAAM,CAAC,KAAK,EAAE,CAAC;gBAAC,CAAC;YACzF,CAAC,CAAC,CAAC;QACL,CAAC,EAAE,IAAI,CAAC,CAAC;QACT,SAAS,CAAC,KAAK,EAAE,CAAC;QAClB,KAAK,CAAC,KAAK,IAAI,EAAE;YACf,IAAI,CAAC;gBACH,IAAI,KAAK,EAAE,MAAM,OAAO,IAAI,MAAM,CAAC,QAAQ,EAAE,CAAC;oBAC5C,IAAI,MAAM;wBAAE,MAAM;oBAClB,IAAI,OAAO,CAAC,YAAY,KAAK,oBAAoB;wBAAE,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,SAAS,EAAE,SAAS,EAAE,MAAM,CAAC,OAAO,CAAC,EAAE,IAAI,EAAE,MAAM,CAAC,OAAO,CAAC,IAAI,IAAI,EAAE,CAAC,EAAE,CAAC,CAAC;yBAC7I,IAAI,OAAO,CAAC,YAAY,KAAK,sBAAsB,EAAE,CAAC;wBACzD,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,eAAe,EAAE,SAAS,EAAE,MAAM,CAAC,OAAO,EAAE,CAAC,EAAE,IAAI,EAAE,MAAM,CAAC,OAAO,CAAC,IAAI,IAAI,EAAE,CAAC,EAAE,CAAC,CAAC;wBACvG,IAAI,QAAQ,EAAE,CAAC;4BAAC,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,QAAQ,EAAE,CAAC,CAAC;4BAAC,QAAQ,GAAG,SAAS,CAAC;wBAAC,CAAC;oBACzF,CAAC;yBAAM,IAAI,OAAO,CAAC,KAAK;wBAAE,MAAM,IAAA,2BAAkB,EAAC,KAAK,EAAE,SAAS,EAAE,OAAO,CAAC,CAAC;gBAChF,CAAC;gBACD,IAAI,CAAC,MAAM;oBAAE,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,EAAE,uBAAuB,EAAE,SAAS,EAAE,IAAI,EAAE,CAAC,CAAC;gBAC5F,MAAM,CAAC,KAAK,EAAE,CAAC;YACjB,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBAAC,MAAM,CAAC,KAAK,CAAC,KAAK,YAAY,kBAAU,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,kBAAU,CAAC,oBAAoB,CAAC,CAAC,CAAC;gBAAC,MAAM,CAAC,KAAK,EAAE,CAAC;YAAC,CAAC;oBACrH,CAAC;gBAAC,MAAM,GAAG,IAAI,CAAC;gBAAC,aAAa,CAAC,SAAS,CAAC,CAAC;YAAC,CAAC;QACtD,CAAC,CAAC,EAAE,CAAC;QACL,OAAO;YACL,MAAM;YACN,SAAS,EAAE,SAAS;YACpB,aAAa,EAAE,KAAK,EAAC,EAAE,EAAC,EAAE;gBACxB,IAAI,QAAQ;oBAAE,MAAM,IAAI,kBAAU,CAAC,oBAAoB,CAAC,CAAC;gBACzD,QAAQ,GAAG,EAAE,CAAC;gBACd,MAAM,MAAM,CAAC,IAAI,CAAC,EAAE,YAAY,EAAE,mBAAmB,EAAE,aAAa,EAAE,EAAE,EAAE,MAAM,EAAE,IAAI,EAAE,WAAW,EAAE,OAAO,CAAC,MAAM,CAAC,UAAU,EAAE,CAAC,CAAC;YACpI,CAAC;YACD,KAAK,EAAE,KAAK,IAAI,EAAE,GAAG,MAAM,GAAG,IAAI,CAAC,CAAC,aAAa,CAAC,SAAS,CAAC,CAAC,CAAC,MAAM,CAAC,KAAK,EAAE,CAAC,CAAC,MAAM,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC;SAChG,CAAC;IACJ,CAAC;CACF;AArDD,sCAqDC"}

package/dist/voice/providers/elevenlabs-tts.d.ts

@@ -0,0 +1,26 @@
+import { TtsProvider, TtsAudioChunk } from '../types';
+import { SocketFactory } from './socket';
+export declare class ElevenLabsTts implements TtsProvider {
+    private readonly key;
+    private readonly model;
+    private readonly connect;
+    readonly id: string;
+    readonly capabilities: {
+        textStreaming: boolean;
+        wordAlignment: boolean;
+        outputFormats: import("../types").AudioFormat[];
+    };
+    constructor(key: string, model?: string, connect?: SocketFactory, id?: string);
+    synthesize(options: Parameters<TtsProvider['synthesize']>[0]): AsyncIterable<TtsAudioChunk>;
+    synthesizeFile(options: {
+        text: string;
+        voiceId: string;
+        signal: AbortSignal;
+    }): Promise<{
+        bytes: Uint8Array;
+        mime: 'audio/mpeg';
+        name: string;
+    }>;
+    private audio;
+}
+//# sourceMappingURL=elevenlabs-tts.d.ts.map

package/dist/voice/providers/elevenlabs-tts.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"elevenlabs-tts.d.ts","sourceRoot":"","sources":["../../../src/voice/providers/elevenlabs-tts.ts"],"names":[],"mappings":"AACA,OAAO,EAAS,WAAW,EAAc,aAAa,EAAuB,MAAM,UAAU,CAAC;AAC9F,OAAO,EAAsB,aAAa,EAAE,MAAM,UAAU,CAAC;AAE7D,qBAAa,aAAc,YAAW,WAAW;IAEnC,OAAO,CAAC,QAAQ,CAAC,GAAG;IAAU,OAAO,CAAC,QAAQ,CAAC,KAAK;IAA+B,OAAO,CAAC,QAAQ,CAAC,OAAO;IAAsC,QAAQ,CAAC,EAAE,EAAE,MAAM;IADhL,QAAQ,CAAC,YAAY;;;;MAAyE;gBACjE,GAAG,EAAE,MAAM,EAAmB,KAAK,SAA6B,EAAmB,OAAO,GAAE,aAAkC,EAAW,EAAE,GAAE,MAAqB;IACxL,UAAU,CAAC,OAAO,EAAE,UAAU,CAAC,WAAW,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,aAAa,CAAC,aAAa,CAAC;IAM5F,cAAc,CAAC,OAAO,EAAE;QAAE,IAAI,EAAE,MAAM,CAAC;QAAC,OAAO,EAAE,MAAM,CAAC;QAAC,MAAM,EAAE,WAAW,CAAA;KAAE,GAAG,OAAO,CAAC;QAAE,KAAK,EAAE,UAAU,CAAC;QAAC,IAAI,EAAE,YAAY,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE,CAAC;YASxI,KAAK;CA6BrB"}

package/dist/voice/providers/elevenlabs-tts.js

@@ -0,0 +1,82 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.ElevenLabsTts = void 0;
+const errors_1 = require("../errors");
+const types_1 = require("../types");
+const socket_1 = require("./socket");
+class ElevenLabsTts {
+    constructor(key, model = 'eleven_v3_conversational', connect = socket_1.openProviderSocket, id = 'elevenlabs') {
+        this.key = key;
+        this.model = model;
+        this.connect = connect;
+        this.id = id;
+        this.capabilities = { textStreaming: true, wordAlignment: false, outputFormats: [types_1.PCM16] };
+    }
+    async *synthesize(options) {
+        (0, types_1.validateAudioFormat)(options.outputFormat);
+        if (options.outputFormat.sampleRate !== 16000)
+            throw new types_1.VoiceError('UNSUPPORTED_AUDIO_FORMAT');
+        let chunkSeq = 0;
+        for await (const bytes of this.audio(options, 'pcm_16000'))
+            yield { bytes, format: options.outputFormat, chunkSeq: chunkSeq++ };
+    }
+    async synthesizeFile(options) {
+        const chunks = [];
+        let size = 0;
+        for await (const bytes of this.audio({ ...options, text: (async function* () { yield options.text; })() }, 'mp3_44100_128')) {
+            size += bytes.length;
+            if (size > 5 * 1024 * 1024)
+                throw new types_1.VoiceError('TTS_AUDIO_TOO_LARGE');
+            chunks.push(bytes);
+        }
+        if (!size || options.signal.aborted)
+            throw new types_1.VoiceError('TTS_INCOMPLETE');
+        return { bytes: Buffer.concat(chunks), mime: 'audio/mpeg', name: 'reply.mp3' };
+    }
+    async *audio(options, format) {
+        if (!this.key)
+            throw new types_1.VoiceError('TTS_CREDENTIALS_MISSING');
+        const dialogue = this.model === 'eleven_v3_conversational' || this.model === 'eleven_v3';
+        const url = new URL(dialogue ? 'wss://api.elevenlabs.io/v1/text-to-dialogue/stream-input' : `wss://api.elevenlabs.io/v1/text-to-speech/${encodeURIComponent(options.voiceId)}/stream-input`);
+        url.searchParams.set('seed', '42');
+        url.searchParams.set('model_id', this.model);
+        url.searchParams.set('output_format', format);
+        if (options.language)
+            url.searchParams.set('language_code', options.language);
+        const socket = await this.connect(url.toString(), { 'xi-api-key': this.key }, options.signal);
+        let sendError, finished = false;
+        const sending = (async () => {
+            await socket.send(dialogue ? { voices: [options.voiceId], voice_settings: { stability: 1 } } : { text: ' ', voice_settings: { stability: 0.5 }, generation_config: { chunk_length_schedule: [50, 120, 160, 290] } });
+            for await (const text of options.text) {
+                if (options.signal.aborted)
+                    return;
+                await socket.send(dialogue ? { inputs: [{ text, voice_id: options.voiceId }] } : { text, try_trigger_generation: true });
+            }
+            await socket.send(dialogue ? { close_socket: true } : { text: '' });
+        })().catch(error => { sendError = error; socket.close(); });
+        try {
+            for await (const message of socket.messages) {
+                if (options.signal.aborted)
+                    return;
+                if (message.error)
+                    throw (0, errors_1.providerVoiceError)('TTS', undefined, message);
+                if (typeof message.audio === 'string')
+                    yield Buffer.from(message.audio, 'base64');
+                if (message.is_final === true || message.isFinal === true) {
+                    finished = true;
+                    break;
+                }
+            }
+            await sending;
+            if (sendError)
+                throw sendError;
+            if (!finished && !options.signal.aborted)
+                throw new types_1.VoiceError('TTS_INCOMPLETE');
+        }
+        finally {
+            socket.close();
+        }
+    }
+}
+exports.ElevenLabsTts = ElevenLabsTts;
+//# sourceMappingURL=elevenlabs-tts.js.map

package/dist/voice/providers/elevenlabs-tts.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"elevenlabs-tts.js","sourceRoot":"","sources":["../../../src/voice/providers/elevenlabs-tts.ts"],"names":[],"mappings":";;;AAAA,sCAA+C;AAC/C,oCAA8F;AAC9F,qCAA6D;AAE7D,MAAa,aAAa;IAExB,YAA6B,GAAW,EAAmB,QAAQ,0BAA0B,EAAmB,UAAyB,2BAAkB,EAAW,KAAa,YAAY;QAAlK,QAAG,GAAH,GAAG,CAAQ;QAAmB,UAAK,GAAL,KAAK,CAA6B;QAAmB,YAAO,GAAP,OAAO,CAAoC;QAAW,OAAE,GAAF,EAAE,CAAuB;QADtL,iBAAY,GAAG,EAAE,aAAa,EAAE,IAAI,EAAE,aAAa,EAAE,KAAK,EAAE,aAAa,EAAE,CAAC,aAAK,CAAC,EAAE,CAAC;IACoG,CAAC;IACnM,KAAK,CAAC,CAAC,UAAU,CAAC,OAAiD;QACjE,IAAA,2BAAmB,EAAC,OAAO,CAAC,YAAY,CAAC,CAAC;QAC1C,IAAI,OAAO,CAAC,YAAY,CAAC,UAAU,KAAK,KAAK;YAAE,MAAM,IAAI,kBAAU,CAAC,0BAA0B,CAAC,CAAC;QAChG,IAAI,QAAQ,GAAG,CAAC,CAAC;QACjB,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,CAAC,KAAK,CAAC,OAAO,EAAE,WAAW,CAAC;YAAE,MAAM,EAAE,KAAK,EAAE,MAAM,EAAE,OAAO,CAAC,YAAY,EAAE,QAAQ,EAAE,QAAQ,EAAE,EAAE,CAAC;IAClI,CAAC;IACD,KAAK,CAAC,cAAc,CAAC,OAA+D;QAClF,MAAM,MAAM,GAAa,EAAE,CAAC;QAAC,IAAI,IAAI,GAAG,CAAC,CAAC;QAC1C,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,CAAC,KAAK,CAAC,EAAE,GAAG,OAAO,EAAE,IAAI,EAAE,CAAC,KAAK,SAAS,CAAC,MAAM,MAAM,OAAO,CAAC,IAAI,CAAC,CAAC,CAAC,CAAC,EAAE,EAAE,EAAE,eAAe,CAAC,EAAE,CAAC;YAC5H,IAAI,IAAI,KAAK,CAAC,MAAM,CAAC;YAAC,IAAI,IAAI,GAAG,CAAC,GAAG,IAAI,GAAG,IAAI;gBAAE,MAAM,IAAI,kBAAU,CAAC,qBAAqB,CAAC,CAAC;YAC9F,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACrB,CAAC;QACD,IAAI,CAAC,IAAI,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO;YAAE,MAAM,IAAI,kBAAU,CAAC,gBAAgB,CAAC,CAAC;QAC5E,OAAO,EAAE,KAAK,EAAE,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,WAAW,EAAE,CAAC;IACjF,CAAC;IACO,KAAK,CAAC,CAAC,KAAK,CAAC,OAAiG,EAAE,MAAc;QACpI,IAAI,CAAC,IAAI,CAAC,GAAG;YAAE,MAAM,IAAI,kBAAU,CAAC,yBAAyB,CAAC,CAAC;QAC/D,MAAM,QAAQ,GAAG,IAAI,CAAC,KAAK,KAAK,0BAA0B,IAAI,IAAI,CAAC,KAAK,KAAK,WAAW,CAAC;QACzF,MAAM,GAAG,GAAG,IAAI,GAAG,CAAC,QAAQ,CAAC,CAAC,CAAC,0DAA0D,CAAC,CAAC,CAAC,6CAA6C,kBAAkB,CAAC,OAAO,CAAC,OAAO,CAAC,eAAe,CAAC,CAAC;QAC7L,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;QACnC,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,UAAU,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC;QAAC,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,eAAe,EAAE,MAAM,CAAC,CAAC;QAC5F,IAAI,OAAO,CAAC,QAAQ;YAAE,GAAG,CAAC,YAAY,CAAC,GAAG,CAAC,eAAe,EAAE,OAAO,CAAC,QAAQ,CAAC,CAAC;QAC9E,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC,QAAQ,EAAE,EAAE,EAAE,YAAY,EAAE,IAAI,CAAC,GAAG,EAAE,EAAE,OAAO,CAAC,MAAM,CAAC,CAAC;QAC9F,IAAI,SAAkB,EAAE,QAAQ,GAAG,KAAK,CAAC;QACzC,MAAM,OAAO,GAAG,CAAC,KAAK,IAAI,EAAE;YAC1B,MAAM,MAAM,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,OAAO,CAAC,OAAO,CAAC,EAAE,cAAc,EAAE,EAAE,SAAS,EAAE,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,GAAG,EAAE,cAAc,EAAE,EAAE,SAAS,EAAE,GAAG,EAAE,EAAE,iBAAiB,EAAE,EAAE,qBAAqB,EAAE,CAAC,EAAE,EAAC,GAAG,EAAC,GAAG,EAAC,GAAG,CAAC,EAAE,EAAE,CAAC,CAAC;YAClN,IAAI,KAAK,EAAE,MAAM,IAAI,IAAI,OAAO,CAAC,IAAI,EAAE,CAAC;gBACtC,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO;oBAAE,OAAO;gBACnC,MAAM,MAAM,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,CAAC,EAAE,IAAI,EAAE,QAAQ,EAAE,OAAO,CAAC,OAAO,EAAE,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,sBAAsB,EAAE,IAAI,EAAE,CAAC,CAAC;YAC3H,CAAC;YACD,MAAM,MAAM,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,YAAY,EAAE,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE,IAAI,EAAE,EAAE,EAAE,CAAC,CAAC;QACtE,CAAC,CAAC,EAAE,CAAC,KAAK,CAAC,KAAK,CAAC,EAAE,GAAG,SAAS,GAAG,KAAK,CAAC,CAAC,MAAM,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC;QAC5D,IAAI,CAAC;YACH,IAAI,KAAK,EAAE,MAAM,OAAO,IAAI,MAAM,CAAC,QAAQ,EAAE,CAAC;gBAC5C,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO;oBAAE,OAAO;gBACnC,IAAI,OAAO,CAAC,KAAK;oBAAE,MAAM,IAAA,2BAAkB,EAAC,KAAK,EAAE,SAAS,EAAE,OAAO,CAAC,CAAC;gBACvE,IAAI,OAAO,OAAO,CAAC,KAAK,KAAK,QAAQ;oBAAE,MAAM,MAAM,CAAC,IAAI,CAAC,OAAO,CAAC,KAAK,EAAE,QAAQ,CAAC,CAAC;gBAClF,IAAI,OAAO,CAAC,QAAQ,KAAK,IAAI,IAAI,OAAO,CAAC,OAAO,KAAK,IAAI,EAAE,CAAC;oBAAC,QAAQ,GAAG,IAAI,CAAC;oBAAC,MAAM;gBAAC,CAAC;YACxF,CAAC;YACD,MAAM,OAAO,CAAC;YACd,IAAI,SAAS;gBAAE,MAAM,SAAS,CAAC;YAC/B,IAAI,CAAC,QAAQ,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,OAAO;gBAAE,MAAM,IAAI,kBAAU,CAAC,gBAAgB,CAAC,CAAC;QACnF,CAAC;gBAAS,CAAC;YAAC,MAAM,CAAC,KAAK,EAAE,CAAC;QAAC,CAAC;IAC/B,CAAC;CACF;AA/CD,sCA+CC"}

package/dist/voice/providers/encoded-audio-tts.d.ts

@@ -0,0 +1,21 @@
+import { TtsProvider } from '../types';
+/** Shared bounded MP3-to-PCM streaming for encoded-audio TTS providers. */
+export declare abstract class EncodedAudioTts implements TtsProvider {
+    readonly id: string;
+    abstract readonly capabilities: TtsProvider['capabilities'];
+    constructor(id: string);
+    protected get pcmSampleRate(): number | undefined;
+    private get inputArgs();
+    synthesize(options: Parameters<TtsProvider['synthesize']>[0]): AsyncGenerator<{
+        bytes: Buffer<ArrayBuffer>;
+        format: import("../types").AudioFormat;
+        chunkSeq: number;
+    }, void, unknown>;
+    synthesizeFile(options: Parameters<NonNullable<TtsProvider['synthesizeFile']>>[0]): Promise<{
+        bytes: Buffer<ArrayBufferLike>;
+        mime: "audio/mpeg";
+        name: string;
+    }>;
+    protected abstract audio(options: Pick<Parameters<TtsProvider['synthesize']>[0], 'text' | 'voiceId' | 'signal'>): AsyncIterable<Buffer>;
+}
+//# sourceMappingURL=encoded-audio-tts.d.ts.map

package/dist/voice/providers/encoded-audio-tts.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"encoded-audio-tts.d.ts","sourceRoot":"","sources":["../../../src/voice/providers/encoded-audio-tts.ts"],"names":[],"mappings":"AAEA,OAAO,EAAS,WAAW,EAAmC,MAAM,UAAU,CAAC;AAE/E,2EAA2E;AAC3E,8BAAsB,eAAgB,YAAW,WAAW;IAE9C,QAAQ,CAAC,EAAE,EAAE,MAAM;IAD/B,QAAQ,CAAC,QAAQ,CAAC,YAAY,EAAE,WAAW,CAAC,cAAc,CAAC,CAAC;gBACvC,EAAE,EAAE,MAAM;IAC/B,SAAS,KAAK,aAAa,IAAI,MAAM,GAAG,SAAS,CAEhD;IACD,OAAO,KAAK,SAAS,GAIpB;IACM,UAAU,CAAC,OAAO,EAAE,UAAU,CAAC,WAAW,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC;;;;;IAwF7D,cAAc,CAClB,OAAO,EAAE,UAAU,CAAC,WAAW,CAAC,WAAW,CAAC,gBAAgB,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC;;;;;IAoEpE,SAAS,CAAC,QAAQ,CAAC,KAAK,CACtB,OAAO,EAAE,IAAI,CACX,UAAU,CAAC,WAAW,CAAC,YAAY,CAAC,CAAC,CAAC,CAAC,CAAC,EACxC,MAAM,GAAG,SAAS,GAAG,QAAQ,CAC9B,GACA,aAAa,CAAC,MAAM,CAAC;CACzB"}

package/dist/voice/providers/encoded-audio-tts.js

@@ -0,0 +1,175 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.EncodedAudioTts = void 0;
+const child_process_1 = require("child_process");
+const events_1 = require("events");
+const types_1 = require("../types");
+/** Shared bounded MP3-to-PCM streaming for encoded-audio TTS providers. */
+class EncodedAudioTts {
+    constructor(id) {
+        this.id = id;
+    }
+    get pcmSampleRate() {
+        return undefined;
+    }
+    get inputArgs() {
+        return this.pcmSampleRate
+            ? ['-f', 's16le', '-ar', String(this.pcmSampleRate), '-ac', '1']
+            : ['-f', 'mp3'];
+    }
+    async *synthesize(options) {
+        (0, types_1.validateAudioFormat)(options.outputFormat);
+        if (options.outputFormat.sampleRate !== 16000)
+            throw new types_1.VoiceError('UNSUPPORTED_AUDIO_FORMAT');
+        // Normalize the provider audio to the voice transport format.
+        const abort = new AbortController();
+        const signal = AbortSignal.any([options.signal, abort.signal]);
+        const decoder = (0, child_process_1.spawn)('ffmpeg', [
+            '-nostdin',
+            '-hide_banner',
+            '-loglevel',
+            'error',
+            '-probesize',
+            '32768',
+            '-analyzeduration',
+            '0',
+            ...this.inputArgs,
+            '-i',
+            'pipe:0',
+            '-f',
+            's16le',
+            '-ac',
+            '1',
+            '-ar',
+            '16000',
+            'pipe:1',
+        ], { stdio: ['pipe', 'pipe', 'ignore'] });
+        let failure;
+        decoder.on('error', () => {
+            failure = new types_1.VoiceError('TTS_DECODER_UNAVAILABLE');
+            abort.abort();
+        });
+        decoder.stdin.on('error', () => {
+            failure ?? (failure = new types_1.VoiceError('TTS_DECODE_FAILED'));
+            abort.abort();
+        });
+        const exited = new Promise((resolve) => decoder.once('close', (code) => {
+            if (code && !options.signal.aborted)
+                failure ?? (failure = new types_1.VoiceError('TTS_DECODE_FAILED'));
+            resolve();
+        }));
+        const stop = () => {
+            decoder.kill('SIGKILL');
+        };
+        signal.addEventListener('abort', stop, { once: true });
+        const feeding = (async () => {
+            for await (const bytes of this.audio({ ...options, signal })) {
+                if (!decoder.stdin.write(bytes))
+                    await (0, events_1.once)(decoder.stdin, 'drain', { signal });
+            }
+            decoder.stdin.end();
+        })().catch((error) => {
+            failure ?? (failure = error);
+            abort.abort();
+        });
+        let chunkSeq = 0, pending = Buffer.alloc(0);
+        try {
+            for await (const chunk of decoder.stdout) {
+                const bytes = Buffer.concat([pending, chunk]);
+                const length = bytes.length - (bytes.length % 2);
+                pending = bytes.subarray(length);
+                if (length)
+                    yield {
+                        bytes: bytes.subarray(0, length),
+                        format: options.outputFormat,
+                        chunkSeq: chunkSeq++,
+                    };
+            }
+            await feeding;
+            await exited;
+            if (failure)
+                throw failure;
+            if (!chunkSeq && !options.signal.aborted)
+                throw new types_1.VoiceError('TTS_INCOMPLETE');
+        }
+        finally {
+            abort.abort();
+            signal.removeEventListener('abort', stop);
+            decoder.kill('SIGKILL');
+            await feeding;
+            await exited;
+        }
+    }
+    async synthesizeFile(options) {
+        const chunks = [];
+        let size = 0;
+        for await (const chunk of this.audio({
+            ...options,
+            text: (async function* () {
+                yield options.text;
+            })(),
+        })) {
+            size += chunk.length;
+            if (size > 5 * 1024 * 1024)
+                throw new types_1.VoiceError('TTS_AUDIO_TOO_LARGE');
+            chunks.push(chunk);
+        }
+        if (!size || options.signal.aborted)
+            throw new types_1.VoiceError('TTS_INCOMPLETE');
+        let bytes = Buffer.concat(chunks);
+        if (this.pcmSampleRate) {
+            if (bytes.length % 2)
+                throw new types_1.VoiceError('TTS_INCOMPLETE');
+            bytes = await new Promise((resolve, reject) => {
+                const encoder = (0, child_process_1.spawn)('ffmpeg', [
+                    '-nostdin',
+                    '-hide_banner',
+                    '-loglevel',
+                    'error',
+                    ...this.inputArgs,
+                    '-i',
+                    'pipe:0',
+                    '-f',
+                    'mp3',
+                    'pipe:1',
+                ], { stdio: ['pipe', 'pipe', 'ignore'] });
+                const stop = () => {
+                    encoder.kill('SIGKILL');
+                };
+                const output = [];
+                let length = 0, failure;
+                options.signal.addEventListener('abort', stop, { once: true });
+                encoder.on('error', () => {
+                    failure = new types_1.VoiceError('TTS_DECODER_UNAVAILABLE');
+                });
+                encoder.stdin.on('error', () => {
+                    failure ?? (failure = new types_1.VoiceError('TTS_DECODE_FAILED'));
+                    stop();
+                });
+                encoder.stdout.on('data', (chunk) => {
+                    length += chunk.length;
+                    if (length > 5 * 1024 * 1024) {
+                        failure = new types_1.VoiceError('TTS_AUDIO_TOO_LARGE');
+                        stop();
+                    }
+                    else
+                        output.push(chunk);
+                });
+                encoder.once('close', (code) => {
+                    options.signal.removeEventListener('abort', stop);
+                    if (failure || code || options.signal.aborted)
+                        reject(failure ?? new types_1.VoiceError('TTS_INCOMPLETE'));
+                    else
+                        resolve(Buffer.concat(output));
+                });
+                if (options.signal.aborted)
+                    stop();
+                else
+                    encoder.stdin.end(bytes);
+            });
+        }
+        return { bytes, mime: 'audio/mpeg', name: 'reply.mp3' };
+    }
+}
+exports.EncodedAudioTts = EncodedAudioTts;
+//# sourceMappingURL=encoded-audio-tts.js.map

package/dist/voice/providers/encoded-audio-tts.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"encoded-audio-tts.js","sourceRoot":"","sources":["../../../src/voice/providers/encoded-audio-tts.ts"],"names":[],"mappings":";;;AAAA,iDAAsC;AACtC,mCAA8B;AAC9B,oCAA+E;AAE/E,2EAA2E;AAC3E,MAAsB,eAAe;IAEnC,YAAqB,EAAU;QAAV,OAAE,GAAF,EAAE,CAAQ;IAAG,CAAC;IACnC,IAAc,aAAa;QACzB,OAAO,SAAS,CAAC;IACnB,CAAC;IACD,IAAY,SAAS;QACnB,OAAO,IAAI,CAAC,aAAa;YACvB,CAAC,CAAC,CAAC,IAAI,EAAE,OAAO,EAAE,KAAK,EAAE,MAAM,CAAC,IAAI,CAAC,aAAa,CAAC,EAAE,KAAK,EAAE,GAAG,CAAC;YAChE,CAAC,CAAC,CAAC,IAAI,EAAE,KAAK,CAAC,CAAC;IACpB,CAAC;IACD,KAAK,CAAC,CAAC,UAAU,CAAC,OAAiD;QACjE,IAAA,2BAAmB,EAAC,OAAO,CAAC,YAAY,CAAC,CAAC;QAC1C,IAAI,OAAO,CAAC,YAAY,CAAC,UAAU,KAAK,KAAK;YAC3C,MAAM,IAAI,kBAAU,CAAC,0BAA0B,CAAC,CAAC;QACnD,8DAA8D;QAC9D,MAAM,KAAK,GAAG,IAAI,eAAe,EAAE,CAAC;QACpC,MAAM,MAAM,GAAG,WAAW,CAAC,GAAG,CAAC,CAAC,OAAO,CAAC,MAAM,EAAE,KAAK,CAAC,MAAM,CAAC,CAAC,CAAC;QAC/D,MAAM,OAAO,GAAG,IAAA,qBAAK,EACnB,QAAQ,EACR;YACE,UAAU;YACV,cAAc;YACd,WAAW;YACX,OAAO;YACP,YAAY;YACZ,OAAO;YACP,kBAAkB;YAClB,GAAG;YACH,GAAG,IAAI,CAAC,SAAS;YACjB,IAAI;YACJ,QAAQ;YACR,IAAI;YACJ,OAAO;YACP,KAAK;YACL,GAAG;YACH,KAAK;YACL,OAAO;YACP,QAAQ;SACT,EACD,EAAE,KAAK,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,CACtC,CAAC;QACF,IAAI,OAAgB,CAAC;QACrB,OAAO,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;YACvB,OAAO,GAAG,IAAI,kBAAU,CAAC,yBAAyB,CAAC,CAAC;YACpD,KAAK,CAAC,KAAK,EAAE,CAAC;QAChB,CAAC,CAAC,CAAC;QACH,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;YAC7B,OAAO,KAAP,OAAO,GAAK,IAAI,kBAAU,CAAC,mBAAmB,CAAC,EAAC;YAChD,KAAK,CAAC,KAAK,EAAE,CAAC;QAChB,CAAC,CAAC,CAAC;QACH,MAAM,MAAM,GAAG,IAAI,OAAO,CAAO,CAAC,OAAO,EAAE,EAAE,CAC3C,OAAO,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC,IAAI,EAAE,EAAE;YAC7B,IAAI,IAAI,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,OAAO;gBACjC,OAAO,KAAP,OAAO,GAAK,IAAI,kBAAU,CAAC,mBAAmB,CAAC,EAAC;YAClD,OAAO,EAAE,CAAC;QACZ,CAAC,CAAC,CACH,CAAC;QACF,MAAM,IAAI,GAAG,GAAG,EAAE;YAChB,OAAO,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;QAC1B,CAAC,CAAC;QACF,MAAM,CAAC,gBAAgB,CAAC,OAAO,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC;QACvD,MAAM,OAAO,GAAG,CAAC,KAAK,IAAI,EAAE;YAC1B,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,CAAC,KAAK,CAAC,EAAE,GAAG,OAAO,EAAE,MAAM,EAAE,CAAC,EAAE,CAAC;gBAC7D,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,KAAK,CAAC,KAAK,CAAC;oBAC7B,MAAM,IAAA,aAAI,EAAC,OAAO,CAAC,KAAK,EAAE,OAAO,EAAE,EAAE,MAAM,EAAE,CAAC,CAAC;YACnD,CAAC;YACD,OAAO,CAAC,KAAK,CAAC,GAAG,EAAE,CAAC;QACtB,CAAC,CAAC,EAAE,CAAC,KAAK,CAAC,CAAC,KAAK,EAAE,EAAE;YACnB,OAAO,KAAP,OAAO,GAAK,KAAK,EAAC;YAClB,KAAK,CAAC,KAAK,EAAE,CAAC;QAChB,CAAC,CAAC,CAAC;QACH,IAAI,QAAQ,GAAG,CAAC,EACd,OAAO,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;QAC5B,IAAI,CAAC;YACH,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,OAAO,CAAC,MAAM,EAAE,CAAC;gBACzC,MAAM,KAAK,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC,OAAO,EAAE,KAAK,CAAC,CAAC,CAAC;gBAC9C,MAAM,MAAM,GAAG,KAAK,CAAC,MAAM,GAAG,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;gBACjD,OAAO,GAAG,KAAK,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC;gBACjC,IAAI,MAAM;oBACR,MAAM;wBACJ,KAAK,EAAE,KAAK,CAAC,QAAQ,CAAC,CAAC,EAAE,MAAM,CAAC;wBAChC,MAAM,EAAE,OAAO,CAAC,YAAY;wBAC5B,QAAQ,EAAE,QAAQ,EAAE;qBACrB,CAAC;YACN,CAAC;YACD,MAAM,OAAO,CAAC;YACd,MAAM,MAAM,CAAC;YACb,IAAI,OAAO;gBAAE,MAAM,OAAO,CAAC;YAC3B,IAAI,CAAC,QAAQ,IAAI,CAAC,OAAO,CAAC,MAAM,CAAC,OAAO;gBACtC,MAAM,IAAI,kBAAU,CAAC,gBAAgB,CAAC,CAAC;QAC3C,CAAC;gBAAS,CAAC;YACT,KAAK,CAAC,KAAK,EAAE,CAAC;YACd,MAAM,CAAC,mBAAmB,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;YAC1C,OAAO,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;YACxB,MAAM,OAAO,CAAC;YACd,MAAM,MAAM,CAAC;QACf,CAAC;IACH,CAAC;IACD,KAAK,CAAC,cAAc,CAClB,OAAkE;QAElE,MAAM,MAAM,GAAa,EAAE,CAAC;QAC5B,IAAI,IAAI,GAAG,CAAC,CAAC;QACb,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,CAAC,KAAK,CAAC;YACnC,GAAG,OAAO;YACV,IAAI,EAAE,CAAC,KAAK,SAAS,CAAC;gBACpB,MAAM,OAAO,CAAC,IAAI,CAAC;YACrB,CAAC,CAAC,EAAE;SACL,CAAC,EAAE,CAAC;YACH,IAAI,IAAI,KAAK,CAAC,MAAM,CAAC;YACrB,IAAI,IAAI,GAAG,CAAC,GAAG,IAAI,GAAG,IAAI;gBAAE,MAAM,IAAI,kBAAU,CAAC,qBAAqB,CAAC,CAAC;YACxE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;QACrB,CAAC;QACD,IAAI,CAAC,IAAI,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO;YAAE,MAAM,IAAI,kBAAU,CAAC,gBAAgB,CAAC,CAAC;QAC5E,IAAI,KAAK,GAAW,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;QAC1C,IAAI,IAAI,CAAC,aAAa,EAAE,CAAC;YACvB,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC;gBAAE,MAAM,IAAI,kBAAU,CAAC,gBAAgB,CAAC,CAAC;YAC7D,KAAK,GAAG,MAAM,IAAI,OAAO,CAAS,CAAC,OAAO,EAAE,MAAM,EAAE,EAAE;gBACpD,MAAM,OAAO,GAAG,IAAA,qBAAK,EACnB,QAAQ,EACR;oBACE,UAAU;oBACV,cAAc;oBACd,WAAW;oBACX,OAAO;oBACP,GAAG,IAAI,CAAC,SAAS;oBACjB,IAAI;oBACJ,QAAQ;oBACR,IAAI;oBACJ,KAAK;oBACL,QAAQ;iBACT,EACD,EAAE,KAAK,EAAE,CAAC,MAAM,EAAE,MAAM,EAAE,QAAQ,CAAC,EAAE,CACtC,CAAC;gBACF,MAAM,IAAI,GAAG,GAAG,EAAE;oBAChB,OAAO,CAAC,IAAI,CAAC,SAAS,CAAC,CAAC;gBAC1B,CAAC,CAAC;gBACF,MAAM,MAAM,GAAa,EAAE,CAAC;gBAC5B,IAAI,MAAM,GAAG,CAAC,EACZ,OAAgB,CAAC;gBACnB,OAAO,CAAC,MAAM,CAAC,gBAAgB,CAAC,OAAO,EAAE,IAAI,EAAE,EAAE,IAAI,EAAE,IAAI,EAAE,CAAC,CAAC;gBAC/D,OAAO,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;oBACvB,OAAO,GAAG,IAAI,kBAAU,CAAC,yBAAyB,CAAC,CAAC;gBACtD,CAAC,CAAC,CAAC;gBACH,OAAO,CAAC,KAAK,CAAC,EAAE,CAAC,OAAO,EAAE,GAAG,EAAE;oBAC7B,OAAO,KAAP,OAAO,GAAK,IAAI,kBAAU,CAAC,mBAAmB,CAAC,EAAC;oBAChD,IAAI,EAAE,CAAC;gBACT,CAAC,CAAC,CAAC;gBACH,OAAO,CAAC,MAAM,CAAC,EAAE,CAAC,MAAM,EAAE,CAAC,KAAK,EAAE,EAAE;oBAClC,MAAM,IAAI,KAAK,CAAC,MAAM,CAAC;oBACvB,IAAI,MAAM,GAAG,CAAC,GAAG,IAAI,GAAG,IAAI,EAAE,CAAC;wBAC7B,OAAO,GAAG,IAAI,kBAAU,CAAC,qBAAqB,CAAC,CAAC;wBAChD,IAAI,EAAE,CAAC;oBACT,CAAC;;wBAAM,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;gBAC5B,CAAC,CAAC,CAAC;gBACH,OAAO,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC,IAAI,EAAE,EAAE;oBAC7B,OAAO,CAAC,MAAM,CAAC,mBAAmB,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;oBAClD,IAAI,OAAO,IAAI,IAAI,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO;wBAC3C,MAAM,CAAC,OAAO,IAAI,IAAI,kBAAU,CAAC,gBAAgB,CAAC,CAAC,CAAC;;wBACjD,OAAO,CAAC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,CAAC;gBACtC,CAAC,CAAC,CAAC;gBACH,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO;oBAAE,IAAI,EAAE,CAAC;;oBAC9B,OAAO,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,CAAC,CAAC;YAChC,CAAC,CAAC,CAAC;QACL,CAAC;QACD,OAAO,EAAE,KAAK,EAAE,IAAI,EAAE,YAAqB,EAAE,IAAI,EAAE,WAAW,EAAE,CAAC;IACnE,CAAC;CAOF;AA9KD,0CA8KC"}

package/dist/voice/providers/fake.d.ts

@@ -0,0 +1,39 @@
+import { AudioFormat, SttEvent, SttProvider, SttSession, TtsProvider } from '../types';
+import { BoundedQueue } from '../queue';
+export declare class FakeSttSession implements SttSession {
+    readonly events: BoundedQueue<SttEvent>;
+    readonly frames: Uint8Array[];
+    commitId?: string;
+    pushAudio(frame: Uint8Array): Promise<void>;
+    commitSegment(commitId: string): Promise<void>;
+    emit(event: SttEvent): void;
+    close(): Promise<void>;
+}
+export declare class FakeSttProvider implements SttProvider {
+    readonly id = "fake";
+    readonly capabilities: {
+        partials: boolean;
+        manualCommit: boolean;
+        speechActivityEvents: boolean;
+    };
+    readonly sessions: FakeSttSession[];
+    open(): Promise<FakeSttSession>;
+}
+export declare class FakeTtsProvider implements TtsProvider {
+    readonly id = "fake";
+    readonly capabilities: {
+        textStreaming: boolean;
+        wordAlignment: boolean;
+        outputFormats: AudioFormat[];
+    };
+    synthesize(options: {
+        text: AsyncIterable<string>;
+        outputFormat: AudioFormat;
+        signal: AbortSignal;
+    }): AsyncGenerator<{
+        bytes: Uint8Array<ArrayBuffer>;
+        format: AudioFormat;
+        chunkSeq: number;
+    }, void, unknown>;
+}
+//# sourceMappingURL=fake.d.ts.map

package/dist/voice/providers/fake.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"fake.d.ts","sourceRoot":"","sources":["../../../src/voice/providers/fake.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,WAAW,EAAS,QAAQ,EAAE,WAAW,EAAE,UAAU,EAAE,WAAW,EAAE,MAAM,UAAU,CAAC;AAC9F,OAAO,EAAE,YAAY,EAAE,MAAM,UAAU,CAAC;AAExC,qBAAa,cAAe,YAAW,UAAU;IAC/C,QAAQ,CAAC,MAAM,yBAA0E;IACzF,QAAQ,CAAC,MAAM,EAAE,UAAU,EAAE,CAAM;IACnC,QAAQ,CAAC,EAAE,MAAM,CAAC;IACZ,SAAS,CAAC,KAAK,EAAE,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC;IAC3C,aAAa,CAAC,QAAQ,EAAE,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;IACpD,IAAI,CAAC,KAAK,EAAE,QAAQ,GAAG,IAAI;IACrB,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC;CAC7B;AACD,qBAAa,eAAgB,YAAW,WAAW;IACjD,QAAQ,CAAC,EAAE,UAAU;IACrB,QAAQ,CAAC,YAAY;;;;MAAsE;IAC3F,QAAQ,CAAC,QAAQ,EAAE,cAAc,EAAE,CAAM;IACnC,IAAI,IAAI,OAAO,CAAC,cAAc,CAAC;CACtC;AACD,qBAAa,eAAgB,YAAW,WAAW;IACjD,QAAQ,CAAC,EAAE,UAAU;IACrB,QAAQ,CAAC,YAAY;;;;MAAyE;IACvF,UAAU,CAAC,OAAO,EAAE;QAAE,IAAI,EAAE,aAAa,CAAC,MAAM,CAAC,CAAC;QAAC,YAAY,EAAE,WAAW,CAAC;QAAC,MAAM,EAAE,WAAW,CAAA;KAAE;;;;;CAO3G"}

package/dist/voice/providers/fake.js

@@ -0,0 +1,41 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.FakeTtsProvider = exports.FakeSttProvider = exports.FakeSttSession = void 0;
+const types_1 = require("../types");
+const queue_1 = require("../queue");
+class FakeSttSession {
+    constructor() {
+        this.events = new queue_1.BoundedQueue(65536, item => JSON.stringify(item).length);
+        this.frames = [];
+    }
+    async pushAudio(frame) { this.frames.push(frame.slice()); }
+    async commitSegment(commitId) { this.commitId = commitId; }
+    emit(event) { this.events.push(event); }
+    async close() { this.events.close(); }
+}
+exports.FakeSttSession = FakeSttSession;
+class FakeSttProvider {
+    constructor() {
+        this.id = 'fake';
+        this.capabilities = { partials: true, manualCommit: true, speechActivityEvents: true };
+        this.sessions = [];
+    }
+    async open() { const session = new FakeSttSession(); this.sessions.push(session); return session; }
+}
+exports.FakeSttProvider = FakeSttProvider;
+class FakeTtsProvider {
+    constructor() {
+        this.id = 'fake';
+        this.capabilities = { textStreaming: true, wordAlignment: false, outputFormats: [types_1.PCM16] };
+    }
+    async *synthesize(options) {
+        let chunkSeq = 0;
+        for await (const _text of options.text) {
+            if (options.signal.aborted)
+                return;
+            yield { bytes: new Uint8Array(640), format: options.outputFormat, chunkSeq: chunkSeq++ };
+        }
+    }
+}
+exports.FakeTtsProvider = FakeTtsProvider;
+//# sourceMappingURL=fake.js.map

package/dist/voice/providers/fake.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"fake.js","sourceRoot":"","sources":["../../../src/voice/providers/fake.ts"],"names":[],"mappings":";;;AAAA,oCAA8F;AAC9F,oCAAwC;AAExC,MAAa,cAAc;IAA3B;QACW,WAAM,GAAG,IAAI,oBAAY,CAAW,KAAK,EAAE,IAAI,CAAC,EAAE,CAAC,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC,CAAC;QAChF,WAAM,GAAiB,EAAE,CAAC;IAMrC,CAAC;IAJC,KAAK,CAAC,SAAS,CAAC,KAAiB,IAAmB,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC,CAAC;IACtF,KAAK,CAAC,aAAa,CAAC,QAAgB,IAAmB,IAAI,CAAC,QAAQ,GAAG,QAAQ,CAAC,CAAC,CAAC;IAClF,IAAI,CAAC,KAAe,IAAU,IAAI,CAAC,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC;IACxD,KAAK,CAAC,KAAK,KAAoB,IAAI,CAAC,MAAM,CAAC,KAAK,EAAE,CAAC,CAAC,CAAC;CACtD;AARD,wCAQC;AACD,MAAa,eAAe;IAA5B;QACW,OAAE,GAAG,MAAM,CAAC;QACZ,iBAAY,GAAG,EAAE,QAAQ,EAAE,IAAI,EAAE,YAAY,EAAE,IAAI,EAAE,oBAAoB,EAAE,IAAI,EAAE,CAAC;QAClF,aAAQ,GAAqB,EAAE,CAAC;IAE3C,CAAC;IADC,KAAK,CAAC,IAAI,KAA8B,MAAM,OAAO,GAAG,IAAI,cAAc,EAAE,CAAC,CAAC,IAAI,CAAC,QAAQ,CAAC,IAAI,CAAC,OAAO,CAAC,CAAC,CAAC,OAAO,OAAO,CAAC,CAAC,CAAC;CAC7H;AALD,0CAKC;AACD,MAAa,eAAe;IAA5B;QACW,OAAE,GAAG,MAAM,CAAC;QACZ,iBAAY,GAAG,EAAE,aAAa,EAAE,IAAI,EAAE,aAAa,EAAE,KAAK,EAAE,aAAa,EAAE,CAAC,aAAK,CAAC,EAAE,CAAC;IAQhG,CAAC;IAPC,KAAK,CAAC,CAAC,UAAU,CAAC,OAAwF;QACxG,IAAI,QAAQ,GAAG,CAAC,CAAC;QACjB,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,OAAO,CAAC,IAAI,EAAE,CAAC;YACvC,IAAI,OAAO,CAAC,MAAM,CAAC,OAAO;gBAAE,OAAO;YACnC,MAAM,EAAE,KAAK,EAAE,IAAI,UAAU,CAAC,GAAG,CAAC,EAAE,MAAM,EAAE,OAAO,CAAC,YAAY,EAAE,QAAQ,EAAE,QAAQ,EAAE,EAAE,CAAC;QAC3F,CAAC;IACH,CAAC;CACF;AAVD,0CAUC"}

package/dist/voice/providers/gemini-live-stt.d.ts

@@ -0,0 +1,21 @@
+import { SttProvider, SttSession, AudioFormat } from '../types';
+import { SocketFactory } from './socket';
+/** Dedicated transcription, not Gemini's audio-to-audio agent. */
+export declare class GeminiLiveStt implements SttProvider {
+    readonly id: string;
+    private readonly model;
+    private readonly connect;
+    readonly capabilities: {
+        partials: boolean;
+        manualCommit: boolean;
+        speechActivityEvents: boolean;
+        mode: "realtime";
+    };
+    constructor(id: string, model: string, connect?: SocketFactory);
+    open(options: {
+        format: AudioFormat;
+        language?: string;
+        signal: AbortSignal;
+    }): Promise<SttSession>;
+}
+//# sourceMappingURL=gemini-live-stt.d.ts.map

package/dist/voice/providers/gemini-live-stt.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"gemini-live-stt.d.ts","sourceRoot":"","sources":["../../../src/voice/providers/gemini-live-stt.ts"],"names":[],"mappings":"AACA,OAAO,EACL,WAAW,EACX,UAAU,EAEV,WAAW,EAEZ,MAAM,UAAU,CAAC;AAGlB,OAAO,EAAsB,aAAa,EAAE,MAAM,UAAU,CAAC;AAE7D,kEAAkE;AAClE,qBAAa,aAAc,YAAW,WAAW;IAQ7C,QAAQ,CAAC,EAAE,EAAE,MAAM;IACnB,OAAO,CAAC,QAAQ,CAAC,KAAK;IACtB,OAAO,CAAC,QAAQ,CAAC,OAAO;IAT1B,QAAQ,CAAC,YAAY;;;;;MAKnB;gBAES,EAAE,EAAE,MAAM,EACF,KAAK,EAAE,MAAM,EACb,OAAO,GAAE,aAAkC;IAExD,IAAI,CAAC,OAAO,EAAE;QAClB,MAAM,EAAE,WAAW,CAAC;QACpB,QAAQ,CAAC,EAAE,MAAM,CAAC;QAClB,MAAM,EAAE,WAAW,CAAC;KACrB,GAAG,OAAO,CAAC,UAAU,CAAC;CA8IxB"}