ugly-app 0.1.0

package/dist/server/audio/STTStream.d.ts

@@ -0,0 +1,12 @@
+import type { STTAudioMsg, STTErrorMsg, STTReadyMsg, STTStartMsg, STTStopMsg, STTStoppedMsg, STTTranscriptMsg } from '../../shared/Audio.js';
+import type { STTProvider } from './stt/types.js';
+type STTServerMsg = STTReadyMsg | STTTranscriptMsg | STTErrorMsg | STTStoppedMsg;
+type SendFn = (msg: STTServerMsg) => void;
+export declare function createSTTStreamManager(getProvider: () => STTProvider | undefined): {
+    handleStart(msg: STTStartMsg, send: SendFn): Promise<void>;
+    handleAudio(msg: STTAudioMsg): void;
+    handleStop(msg: STTStopMsg): Promise<void>;
+    cleanup(): Promise<void>;
+};
+export {};
+//# sourceMappingURL=STTStream.d.ts.map

package/dist/server/audio/STTStream.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"STTStream.d.ts","sourceRoot":"","sources":["../../../src/server/audio/STTStream.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EACV,WAAW,EACX,WAAW,EACX,WAAW,EACX,WAAW,EACX,UAAU,EACV,aAAa,EACb,gBAAgB,EACjB,MAAM,uBAAuB,CAAC;AAE/B,OAAO,KAAK,EAAE,WAAW,EAAc,MAAM,gBAAgB,CAAC;AAE9D,KAAK,YAAY,GACb,WAAW,GACX,gBAAgB,GAChB,WAAW,GACX,aAAa,CAAC;AAClB,KAAK,MAAM,GAAG,CAAC,GAAG,EAAE,YAAY,KAAK,IAAI,CAAC;AAO1C,wBAAgB,sBAAsB,CACpC,WAAW,EAAE,MAAM,WAAW,GAAG,SAAS;qBAKjB,WAAW,QAAQ,MAAM,GAAG,OAAO,CAAC,IAAI,CAAC;qBAqC/C,WAAW,GAAG,IAAI;oBAQb,UAAU,GAAG,OAAO,CAAC,IAAI,CAAC;eAQ/B,OAAO,CAAC,IAAI,CAAC;EAQjC"}

package/dist/server/audio/STTStream.js

@@ -0,0 +1,59 @@
+import { resample48kTo16k } from './resample.js';
+export function createSTTStreamManager(getProvider) {
+    const activeSessions = new Map();
+    return {
+        async handleStart(msg, send) {
+            const { streamId } = msg;
+            const provider = getProvider();
+            if (!provider) {
+                send({
+                    type: 'stt:error',
+                    streamId,
+                    message: 'No STT provider available. Set GROQ_API_KEY.',
+                });
+                return;
+            }
+            try {
+                const session = await provider.connect((result) => {
+                    send({
+                        type: 'stt:transcript',
+                        streamId,
+                        text: result.text,
+                        isFinal: result.isFinal,
+                        lang: result.lang,
+                        words: result.words,
+                    });
+                }, (err) => {
+                    send({ type: 'stt:error', streamId, message: err });
+                }, msg.lang, { requestWords: msg.requestWords });
+                activeSessions.set(streamId, { session, send });
+                send({ type: 'stt:ready', streamId });
+            }
+            catch (err) {
+                send({ type: 'stt:error', streamId, message: String(err) });
+            }
+        },
+        handleAudio(msg) {
+            const active = activeSessions.get(msg.streamId);
+            if (!active)
+                return;
+            const raw = Buffer.from(msg.audio, 'base64');
+            const resampled = resample48kTo16k(raw);
+            active.session.sendAudio(resampled);
+        },
+        async handleStop(msg) {
+            const active = activeSessions.get(msg.streamId);
+            if (!active)
+                return;
+            activeSessions.delete(msg.streamId);
+            await active.session.stop();
+            active.send({ type: 'stt:stopped', streamId: msg.streamId });
+        },
+        async cleanup() {
+            const entries = [...activeSessions.entries()];
+            activeSessions.clear();
+            await Promise.all(entries.map(([, active]) => active.session.stop().catch(() => { })));
+        },
+    };
+}
+//# sourceMappingURL=STTStream.js.map

package/dist/server/audio/STTStream.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"STTStream.js","sourceRoot":"","sources":["../../../src/server/audio/STTStream.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,gBAAgB,EAAE,MAAM,eAAe,CAAC;AAejD,MAAM,UAAU,sBAAsB,CACpC,WAA0C;IAE1C,MAAM,cAAc,GAAG,IAAI,GAAG,EAAyB,CAAC;IAExD,OAAO;QACL,KAAK,CAAC,WAAW,CAAC,GAAgB,EAAE,IAAY;YAC9C,MAAM,EAAE,QAAQ,EAAE,GAAG,GAAG,CAAC;YACzB,MAAM,QAAQ,GAAG,WAAW,EAAE,CAAC;YAC/B,IAAI,CAAC,QAAQ,EAAE,CAAC;gBACd,IAAI,CAAC;oBACH,IAAI,EAAE,WAAW;oBACjB,QAAQ;oBACR,OAAO,EAAE,8CAA8C;iBACxD,CAAC,CAAC;gBACH,OAAO;YACT,CAAC;YAED,IAAI,CAAC;gBACH,MAAM,OAAO,GAAG,MAAM,QAAQ,CAAC,OAAO,CACpC,CAAC,MAAM,EAAE,EAAE;oBACT,IAAI,CAAC;wBACH,IAAI,EAAE,gBAAgB;wBACtB,QAAQ;wBACR,IAAI,EAAE,MAAM,CAAC,IAAI;wBACjB,OAAO,EAAE,MAAM,CAAC,OAAO;wBACvB,IAAI,EAAE,MAAM,CAAC,IAAI;wBACjB,KAAK,EAAE,MAAM,CAAC,KAAK;qBACpB,CAAC,CAAC;gBACL,CAAC,EACD,CAAC,GAAG,EAAE,EAAE;oBACN,IAAI,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,QAAQ,EAAE,OAAO,EAAE,GAAG,EAAE,CAAC,CAAC;gBACtD,CAAC,EACD,GAAG,CAAC,IAAI,EACR,EAAE,YAAY,EAAE,GAAG,CAAC,YAAY,EAAE,CACnC,CAAC;gBACF,cAAc,CAAC,GAAG,CAAC,QAAQ,EAAE,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC,CAAC;gBAChD,IAAI,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,QAAQ,EAAE,CAAC,CAAC;YACxC,CAAC;YAAC,OAAO,GAAG,EAAE,CAAC;gBACb,IAAI,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,QAAQ,EAAE,OAAO,EAAE,MAAM,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;YAC9D,CAAC;QACH,CAAC;QAED,WAAW,CAAC,GAAgB;YAC1B,MAAM,MAAM,GAAG,cAAc,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;YAChD,IAAI,CAAC,MAAM;gBAAE,OAAO;YACpB,MAAM,GAAG,GAAG,MAAM,CAAC,IAAI,CAAC,GAAG,CAAC,KAAK,EAAE,QAAQ,CAAC,CAAC;YAC7C,MAAM,SAAS,GAAG,gBAAgB,CAAC,GAAG,CAAC,CAAC;YACxC,MAAM,CAAC,OAAO,CAAC,SAAS,CAAC,SAAS,CAAC,CAAC;QACtC,CAAC;QAED,KAAK,CAAC,UAAU,CAAC,GAAe;YAC9B,MAAM,MAAM,GAAG,cAAc,CAAC,GAAG,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;YAChD,IAAI,CAAC,MAAM;gBAAE,OAAO;YACpB,cAAc,CAAC,MAAM,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;YACpC,MAAM,MAAM,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC;YAC5B,MAAM,CAAC,IAAI,CAAC,EAAE,IAAI,EAAE,aAAa,EAAE,QAAQ,EAAE,GAAG,CAAC,QAAQ,EAAE,CAAC,CAAC;QAC/D,CAAC;QAED,KAAK,CAAC,OAAO;YACX,MAAM,OAAO,GAAG,CAAC,GAAG,cAAc,CAAC,OAAO,EAAE,CAAC,CAAC;YAC9C,cAAc,CAAC,KAAK,EAAE,CAAC;YACvB,MAAM,OAAO,CAAC,GAAG,CACf,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,MAAM,CAAC,EAAE,EAAE,CAAC,MAAM,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC,KAAK,CAAC,GAAG,EAAE,GAAE,CAAC,CAAC,CAAC,CACnE,CAAC;QACJ,CAAC;KACF,CAAC;AACJ,CAAC"}

package/dist/server/audio/TTSStream.d.ts

@@ -0,0 +1,16 @@
+import type { TTSChunkMsg, TTSDoneMsg, TTSErrorMsg, TTSStartMsg, TTSWordMsg } from '../../shared/Audio.js';
+import type { TTSProvider } from './tts/types.js';
+export type SendFn = (msg: TTSChunkMsg | TTSWordMsg | TTSDoneMsg | TTSErrorMsg) => void;
+export interface TTSStreamConfig {
+    getProvider(): TTSProvider | undefined;
+}
+/**
+ * Register a stream ID for cancellation. Call when the client sends tts:cancel.
+ */
+export declare function cancelTTSStream(streamId: string): void;
+export interface TTSStreamHandler {
+    handle: (msg: TTSStartMsg, send: SendFn) => Promise<void>;
+    cancel: (streamId: string) => void;
+}
+export declare function createTTSStreamHandler(config: TTSStreamConfig): TTSStreamHandler;
+//# sourceMappingURL=TTSStream.d.ts.map

package/dist/server/audio/TTSStream.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"TTSStream.d.ts","sourceRoot":"","sources":["../../../src/server/audio/TTSStream.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EACV,WAAW,EACX,UAAU,EACV,WAAW,EACX,WAAW,EACX,UAAU,EACX,MAAM,uBAAuB,CAAC;AAE/B,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,gBAAgB,CAAC;AAElD,MAAM,MAAM,MAAM,GAAG,CACnB,GAAG,EAAE,WAAW,GAAG,UAAU,GAAG,UAAU,GAAG,WAAW,KACrD,IAAI,CAAC;AAEV,MAAM,WAAW,eAAe;IAC9B,WAAW,IAAI,WAAW,GAAG,SAAS,CAAC;CACxC;AAaD;;GAEG;AACH,wBAAgB,eAAe,CAAC,QAAQ,EAAE,MAAM,GAAG,IAAI,CAEtD;AAED,MAAM,WAAW,gBAAgB;IAC/B,MAAM,EAAE,CAAC,GAAG,EAAE,WAAW,EAAE,IAAI,EAAE,MAAM,KAAK,OAAO,CAAC,IAAI,CAAC,CAAC;IAC1D,MAAM,EAAE,CAAC,QAAQ,EAAE,MAAM,KAAK,IAAI,CAAC;CACpC;AAED,wBAAgB,sBAAsB,CACpC,MAAM,EAAE,eAAe,GACtB,gBAAgB,CAmKlB"}

package/dist/server/audio/TTSStream.js

@@ -0,0 +1,151 @@
+import { redisGet, redisSet } from '../Redis.js';
+/** Split text into sentences on . ! ? followed by whitespace */
+function splitSentences(text) {
+    const sentences = text.split(/(?<=[.!?])\s+/);
+    return sentences.map((s) => s.trim()).filter((s) => s.length > 0);
+}
+const CACHE_TTL_SECONDS = 7 * 24 * 60 * 60; // 7 days
+/** Active cancelled stream IDs */
+const cancelledStreams = new Set();
+/**
+ * Register a stream ID for cancellation. Call when the client sends tts:cancel.
+ */
+export function cancelTTSStream(streamId) {
+    cancelledStreams.add(streamId);
+}
+export function createTTSStreamHandler(config) {
+    async function handle(msg, send) {
+        const { streamId, text, voice = 'default', requestVisemes } = msg;
+        const provider = config.getProvider();
+        if (!provider) {
+            send({
+                type: 'tts:error',
+                streamId,
+                message: 'No TTS provider available. Set INWORLD_API_KEY.',
+            });
+            return;
+        }
+        const startTime = Date.now();
+        let chunkIndex = 0;
+        let wasCancelled = false;
+        try {
+            const sentences = splitSentences(text);
+            for (const sentence of sentences) {
+                if (cancelledStreams.has(streamId)) {
+                    wasCancelled = true;
+                    break;
+                }
+                const cacheKey = `tts:${provider.name}:${voice}:${sentence}`;
+                let cachedChunks = null;
+                // Try to load from cache
+                try {
+                    const cached = await redisGet(cacheKey);
+                    if (cached) {
+                        cachedChunks = JSON.parse(cached);
+                    }
+                }
+                catch {
+                    // Cache miss or parse error — proceed without cache
+                }
+                if (cachedChunks) {
+                    // Replay from cache
+                    for (const cached of cachedChunks) {
+                        if (cancelledStreams.has(streamId)) {
+                            wasCancelled = true;
+                            break;
+                        }
+                        send({
+                            type: 'tts:chunk',
+                            streamId,
+                            audio: cached.audio,
+                            sampleRate: 24000,
+                            index: chunkIndex++,
+                        });
+                        if (cached.word !== undefined &&
+                            cached.startMs !== undefined &&
+                            cached.durationMs !== undefined) {
+                            send({
+                                type: 'tts:word',
+                                streamId,
+                                word: cached.word,
+                                startMs: cached.startMs,
+                                durationMs: cached.durationMs,
+                            });
+                        }
+                    }
+                    if (wasCancelled)
+                        break;
+                }
+                else {
+                    // Stream from provider and collect for caching
+                    const collectedChunks = [];
+                    for await (const chunk of provider.stream(sentence, voice, {
+                        requestVisemes,
+                    })) {
+                        if (cancelledStreams.has(streamId)) {
+                            wasCancelled = true;
+                            break;
+                        }
+                        const audioB64 = chunk.audio.toString('base64');
+                        send({
+                            type: 'tts:chunk',
+                            streamId,
+                            audio: audioB64,
+                            sampleRate: 24000,
+                            index: chunkIndex++,
+                            visemes: chunk.visemes,
+                        });
+                        if (chunk.word !== undefined &&
+                            chunk.startMs !== undefined &&
+                            chunk.durationMs !== undefined) {
+                            send({
+                                type: 'tts:word',
+                                streamId,
+                                word: chunk.word,
+                                startMs: chunk.startMs,
+                                durationMs: chunk.durationMs,
+                            });
+                        }
+                        collectedChunks.push({
+                            audio: audioB64,
+                            word: chunk.word,
+                            startMs: chunk.startMs,
+                            durationMs: chunk.durationMs,
+                        });
+                    }
+                    if (wasCancelled)
+                        break;
+                    // Cache the collected chunks
+                    if (collectedChunks.length > 0) {
+                        try {
+                            await redisSet(cacheKey, JSON.stringify(collectedChunks), CACHE_TTL_SECONDS);
+                        }
+                        catch {
+                            // Non-fatal: cache write failure is OK
+                        }
+                    }
+                }
+            }
+        }
+        catch (err) {
+            const message = err instanceof Error ? err.message : String(err);
+            send({ type: 'tts:error', streamId, message });
+            return;
+        }
+        finally {
+            cancelledStreams.delete(streamId);
+        }
+        if (!wasCancelled) {
+            send({
+                type: 'tts:done',
+                streamId,
+                durationMs: Date.now() - startTime,
+            });
+        }
+    }
+    return {
+        handle,
+        cancel: (streamId) => cancelledStreams.add(streamId),
+    };
+}
+//# sourceMappingURL=TTSStream.js.map

package/dist/server/audio/TTSStream.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"TTSStream.js","sourceRoot":"","sources":["../../../src/server/audio/TTSStream.ts"],"names":[],"mappings":"AAQA,OAAO,EAAE,QAAQ,EAAE,QAAQ,EAAE,MAAM,aAAa,CAAC;AAWjD,gEAAgE;AAChE,SAAS,cAAc,CAAC,IAAY;IAClC,MAAM,SAAS,GAAG,IAAI,CAAC,KAAK,CAAC,eAAe,CAAC,CAAC;IAC9C,OAAO,SAAS,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;AACpE,CAAC;AAED,MAAM,iBAAiB,GAAG,CAAC,GAAG,EAAE,GAAG,EAAE,GAAG,EAAE,CAAC,CAAC,SAAS;AAErD,kCAAkC;AAClC,MAAM,gBAAgB,GAAG,IAAI,GAAG,EAAU,CAAC;AAE3C;;GAEG;AACH,MAAM,UAAU,eAAe,CAAC,QAAgB;IAC9C,gBAAgB,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC;AACjC,CAAC;AAOD,MAAM,UAAU,sBAAsB,CACpC,MAAuB;IAEvB,KAAK,UAAU,MAAM,CAAC,GAAgB,EAAE,IAAY;QAClD,MAAM,EAAE,QAAQ,EAAE,IAAI,EAAE,KAAK,GAAG,SAAS,EAAE,cAAc,EAAE,GAAG,GAAG,CAAC;QAClE,MAAM,QAAQ,GAAG,MAAM,CAAC,WAAW,EAAE,CAAC;QAEtC,IAAI,CAAC,QAAQ,EAAE,CAAC;YACd,IAAI,CAAC;gBACH,IAAI,EAAE,WAAW;gBACjB,QAAQ;gBACR,OAAO,EAAE,iDAAiD;aAC3D,CAAC,CAAC;YACH,OAAO;QACT,CAAC;QAED,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;QAC7B,IAAI,UAAU,GAAG,CAAC,CAAC;QACnB,IAAI,YAAY,GAAG,KAAK,CAAC;QAEzB,IAAI,CAAC;YACH,MAAM,SAAS,GAAG,cAAc,CAAC,IAAI,CAAC,CAAC;YAEvC,KAAK,MAAM,QAAQ,IAAI,SAAS,EAAE,CAAC;gBACjC,IAAI,gBAAgB,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,CAAC;oBACnC,YAAY,GAAG,IAAI,CAAC;oBACpB,MAAM;gBACR,CAAC;gBAED,MAAM,QAAQ,GAAG,OAAO,QAAQ,CAAC,IAAI,IAAI,KAAK,IAAI,QAAQ,EAAE,CAAC;gBAC7D,IAAI,YAAY,GAKJ,IAAI,CAAC;gBAEjB,yBAAyB;gBACzB,IAAI,CAAC;oBACH,MAAM,MAAM,GAAG,MAAM,QAAQ,CAAC,QAAQ,CAAC,CAAC;oBACxC,IAAI,MAAM,EAAE,CAAC;wBACX,YAAY,GAAG,IAAI,CAAC,KAAK,CAAC,MAAM,CAAC,CAAC;oBACpC,CAAC;gBACH,CAAC;gBAAC,MAAM,CAAC;oBACP,oDAAoD;gBACtD,CAAC;gBAED,IAAI,YAAY,EAAE,CAAC;oBACjB,oBAAoB;oBACpB,KAAK,MAAM,MAAM,IAAI,YAAY,EAAE,CAAC;wBAClC,IAAI,gBAAgB,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,CAAC;4BACnC,YAAY,GAAG,IAAI,CAAC;4BACpB,MAAM;wBACR,CAAC;wBAED,IAAI,CAAC;4BACH,IAAI,EAAE,WAAW;4BACjB,QAAQ;4BACR,KAAK,EAAE,MAAM,CAAC,KAAK;4BACnB,UAAU,EAAE,KAAK;4BACjB,KAAK,EAAE,UAAU,EAAE;yBACpB,CAAC,CAAC;wBAEH,IACE,MAAM,CAAC,IAAI,KAAK,SAAS;4BACzB,MAAM,CAAC,OAAO,KAAK,SAAS;4BAC5B,MAAM,CAAC,UAAU,KAAK,SAAS,EAC/B,CAAC;4BACD,IAAI,CAAC;gCACH,IAAI,EAAE,UAAU;gCAChB,QAAQ;gCACR,IAAI,EAAE,MAAM,CAAC,IAAI;gCACjB,OAAO,EAAE,MAAM,CAAC,OAAO;gCACvB,UAAU,EAAE,MAAM,CAAC,UAAU;6BAC9B,CAAC,CAAC;wBACL,CAAC;oBACH,CAAC;oBACD,IAAI,YAAY;wBAAE,MAAM;gBAC1B,CAAC;qBAAM,CAAC;oBACN,+CAA+C;oBAC/C,MAAM,eAAe,GAKhB,EAAE,CAAC;oBAER,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,QAAQ,CAAC,MAAM,CAAC,QAAQ,EAAE,KAAK,EAAE;wBACzD,cAAc;qBACf,CAAC,EAAE,CAAC;wBACH,IAAI,gBAAgB,CAAC,GAAG,CAAC,QAAQ,CAAC,EAAE,CAAC;4BACnC,YAAY,GAAG,IAAI,CAAC;4BACpB,MAAM;wBACR,CAAC;wBAED,MAAM,QAAQ,GAAG,KAAK,CAAC,KAAK,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAC;wBAEhD,IAAI,CAAC;4BACH,IAAI,EAAE,WAAW;4BACjB,QAAQ;4BACR,KAAK,EAAE,QAAQ;4BACf,UAAU,EAAE,KAAK;4BACjB,KAAK,EAAE,UAAU,EAAE;4BACnB,OAAO,EAAE,KAAK,CAAC,OAAO;yBACvB,CAAC,CAAC;wBAEH,IACE,KAAK,CAAC,IAAI,KAAK,SAAS;4BACxB,KAAK,CAAC,OAAO,KAAK,SAAS;4BAC3B,KAAK,CAAC,UAAU,KAAK,SAAS,EAC9B,CAAC;4BACD,IAAI,CAAC;gCACH,IAAI,EAAE,UAAU;gCAChB,QAAQ;gCACR,IAAI,EAAE,KAAK,CAAC,IAAI;gCAChB,OAAO,EAAE,KAAK,CAAC,OAAO;gCACtB,UAAU,EAAE,KAAK,CAAC,UAAU;6BAC7B,CAAC,CAAC;wBACL,CAAC;wBAED,eAAe,CAAC,IAAI,CAAC;4BACnB,KAAK,EAAE,QAAQ;4BACf,IAAI,EAAE,KAAK,CAAC,IAAI;4BAChB,OAAO,EAAE,KAAK,CAAC,OAAO;4BACtB,UAAU,EAAE,KAAK,CAAC,UAAU;yBAC7B,CAAC,CAAC;oBACL,CAAC;oBAED,IAAI,YAAY;wBAAE,MAAM;oBAExB,6BAA6B;oBAC7B,IAAI,eAAe,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;wBAC/B,IAAI,CAAC;4BACH,MAAM,QAAQ,CACZ,QAAQ,EACR,IAAI,CAAC,SAAS,CAAC,eAAe,CAAC,EAC/B,iBAAiB,CAClB,CAAC;wBACJ,CAAC;wBAAC,MAAM,CAAC;4BACP,uCAAuC;wBACzC,CAAC;oBACH,CAAC;gBACH,CAAC;YACH,CAAC;QACH,CAAC;QAAC,OAAO,GAAY,EAAE,CAAC;YACtB,MAAM,OAAO,GAAG,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;YACjE,IAAI,CAAC,EAAE,IAAI,EAAE,WAAW,EAAE,QAAQ,EAAE,OAAO,EAAE,CAAC,CAAC;YAC/C,OAAO;QACT,CAAC;gBAAS,CAAC;YACT,gBAAgB,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC;QACpC,CAAC;QAED,IAAI,CAAC,YAAY,EAAE,CAAC;YAClB,IAAI,CAAC;gBACH,IAAI,EAAE,UAAU;gBAChB,QAAQ;gBACR,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,SAAS;aACnC,CAAC,CAAC;QACL,CAAC;IACH,CAAC;IAED,OAAO;QACL,MAAM;QACN,MAAM,EAAE,CAAC,QAAgB,EAAE,EAAE,CAAC,gBAAgB,CAAC,GAAG,CAAC,QAAQ,CAAC;KAC7D,CAAC;AACJ,CAAC"}

package/dist/server/audio/index.d.ts

@@ -0,0 +1,5 @@
+export { registerSTTProvider } from './stt/registry.js';
+export { createSTTStreamManager } from './STTStream.js';
+export { registerTTSProvider } from './tts/registry.js';
+export { createTTSStreamHandler } from './TTSStream.js';
+//# sourceMappingURL=index.d.ts.map

package/dist/server/audio/index.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../../src/server/audio/index.ts"],"names":[],"mappings":"AAYA,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AACxD,OAAO,EAAE,sBAAsB,EAAE,MAAM,gBAAgB,CAAC;AACxD,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AACxD,OAAO,EAAE,sBAAsB,EAAE,MAAM,gBAAgB,CAAC"}

package/dist/server/audio/index.js

@@ -0,0 +1,14 @@
+// src/server/audio/index.ts
+// Side-effect: registers all built-in audio providers
+import { groqWhisperSTTProvider } from './stt/GroqWhisper.js';
+import { registerSTTProvider } from './stt/registry.js';
+import { inworldTTSProvider } from './tts/InWorld.js';
+import { registerTTSProvider } from './tts/registry.js';
+// Register providers (last registered = highest priority)
+registerTTSProvider(inworldTTSProvider);
+registerSTTProvider(groqWhisperSTTProvider);
+export { registerSTTProvider } from './stt/registry.js';
+export { createSTTStreamManager } from './STTStream.js';
+export { registerTTSProvider } from './tts/registry.js';
+export { createTTSStreamHandler } from './TTSStream.js';
+//# sourceMappingURL=index.js.map

package/dist/server/audio/index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"index.js","sourceRoot":"","sources":["../../../src/server/audio/index.ts"],"names":[],"mappings":"AAAA,4BAA4B;AAC5B,sDAAsD;AAEtD,OAAO,EAAE,sBAAsB,EAAE,MAAM,sBAAsB,CAAC;AAC9D,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AACxD,OAAO,EAAE,kBAAkB,EAAE,MAAM,kBAAkB,CAAC;AACtD,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AAExD,0DAA0D;AAC1D,mBAAmB,CAAC,kBAAkB,CAAC,CAAC;AACxC,mBAAmB,CAAC,sBAAsB,CAAC,CAAC;AAE5C,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AACxD,OAAO,EAAE,sBAAsB,EAAE,MAAM,gBAAgB,CAAC;AACxD,OAAO,EAAE,mBAAmB,EAAE,MAAM,mBAAmB,CAAC;AACxD,OAAO,EAAE,sBAAsB,EAAE,MAAM,gBAAgB,CAAC"}

package/dist/server/audio/resample.d.ts

@@ -0,0 +1,7 @@
+/**
+ * Downsample PCM16 from 48kHz to 16kHz (3:1 ratio) by simple decimation.
+ * Input: Buffer of 16-bit signed integers at 48kHz mono.
+ * Output: Buffer of 16-bit signed integers at 16kHz mono.
+ */
+export declare function resample48kTo16k(input: Buffer): Buffer;
+//# sourceMappingURL=resample.d.ts.map

package/dist/server/audio/resample.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"resample.d.ts","sourceRoot":"","sources":["../../../src/server/audio/resample.ts"],"names":[],"mappings":"AACA;;;;GAIG;AACH,wBAAgB,gBAAgB,CAAC,KAAK,EAAE,MAAM,GAAG,MAAM,CAUtD"}

package/dist/server/audio/resample.js

@@ -0,0 +1,17 @@
+// src/server/audio/resample.ts
+/**
+ * Downsample PCM16 from 48kHz to 16kHz (3:1 ratio) by simple decimation.
+ * Input: Buffer of 16-bit signed integers at 48kHz mono.
+ * Output: Buffer of 16-bit signed integers at 16kHz mono.
+ */
+export function resample48kTo16k(input) {
+    const inputSamples = input.length / 2; // 16-bit = 2 bytes per sample
+    const outputSamples = Math.floor(inputSamples / 3);
+    const output = Buffer.allocUnsafe(outputSamples * 2);
+    for (let i = 0; i < outputSamples; i++) {
+        const srcIdx = i * 3 * 2;
+        output.writeInt16LE(input.readInt16LE(srcIdx), i * 2);
+    }
+    return output;
+}
+//# sourceMappingURL=resample.js.map

package/dist/server/audio/resample.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"resample.js","sourceRoot":"","sources":["../../../src/server/audio/resample.ts"],"names":[],"mappings":"AAAA,+BAA+B;AAC/B;;;;GAIG;AACH,MAAM,UAAU,gBAAgB,CAAC,KAAa;IAC5C,MAAM,YAAY,GAAG,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC,8BAA8B;IACrE,MAAM,aAAa,GAAG,IAAI,CAAC,KAAK,CAAC,YAAY,GAAG,CAAC,CAAC,CAAC;IACnD,MAAM,MAAM,GAAG,MAAM,CAAC,WAAW,CAAC,aAAa,GAAG,CAAC,CAAC,CAAC;IAErD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,aAAa,EAAE,CAAC,EAAE,EAAE,CAAC;QACvC,MAAM,MAAM,GAAG,CAAC,GAAG,CAAC,GAAG,CAAC,CAAC;QACzB,MAAM,CAAC,YAAY,CAAC,KAAK,CAAC,WAAW,CAAC,MAAM,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,CAAC;IACxD,CAAC;IACD,OAAO,MAAM,CAAC;AAChB,CAAC"}

package/dist/server/audio/stt/GroqWhisper.d.ts

@@ -0,0 +1,3 @@
+import type { STTProvider } from './types.js';
+export declare const groqWhisperSTTProvider: STTProvider;
+//# sourceMappingURL=GroqWhisper.d.ts.map

package/dist/server/audio/stt/GroqWhisper.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"GroqWhisper.d.ts","sourceRoot":"","sources":["../../../../src/server/audio/stt/GroqWhisper.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAEV,WAAW,EAGZ,MAAM,YAAY,CAAC;AAyCpB,eAAO,MAAM,sBAAsB,EAAE,WAwFpC,CAAC"}

package/dist/server/audio/stt/GroqWhisper.js

@@ -0,0 +1,91 @@
+const GROQ_ENDPOINT = 'https://api.groq.com/openai/v1/audio/transcriptions';
+function buildWavHeader(dataLength) {
+    const sampleRate = 16000;
+    const numChannels = 1;
+    const bitsPerSample = 16;
+    const byteRate = sampleRate * numChannels * (bitsPerSample / 8);
+    const blockAlign = numChannels * (bitsPerSample / 8);
+    const buffer = Buffer.alloc(44);
+    buffer.write('RIFF', 0);
+    buffer.writeUInt32LE(36 + dataLength, 4);
+    buffer.write('WAVE', 8);
+    buffer.write('fmt ', 12);
+    buffer.writeUInt32LE(16, 16); // PCM chunk size
+    buffer.writeUInt16LE(1, 20); // PCM format
+    buffer.writeUInt16LE(numChannels, 22);
+    buffer.writeUInt32LE(sampleRate, 24);
+    buffer.writeUInt32LE(byteRate, 28);
+    buffer.writeUInt16LE(blockAlign, 32);
+    buffer.writeUInt16LE(bitsPerSample, 34);
+    buffer.write('data', 36);
+    buffer.writeUInt32LE(dataLength, 40);
+    return buffer;
+}
+export const groqWhisperSTTProvider = {
+    name: 'groq-whisper',
+    apiKeyEnv: 'GROQ_API_KEY',
+    async connect(onTranscript, onError, lang, options) {
+        const apiKey = process.env['GROQ_API_KEY'];
+        if (!apiKey) {
+            throw new Error('GROQ_API_KEY is not set');
+        }
+        const chunks = [];
+        return {
+            sendAudio(pcm16) {
+                chunks.push(pcm16);
+            },
+            async stop() {
+                if (chunks.length === 0) {
+                    return;
+                }
+                const pcmData = Buffer.concat(chunks);
+                const wavHeader = buildWavHeader(pcmData.length);
+                const wavData = Buffer.concat([wavHeader, pcmData]);
+                const formData = new FormData();
+                formData.append('file', new Blob([wavData], { type: 'audio/wav' }), 'audio.wav');
+                formData.append('model', 'whisper-large-v3-turbo');
+                // Use verbose_json when word timestamps are requested — slower but includes word timings
+                // Use json (default) otherwise for lower latency
+                if (options?.requestWords) {
+                    formData.append('response_format', 'verbose_json');
+                    formData.append('timestamp_granularities[]', 'word');
+                }
+                else {
+                    formData.append('response_format', 'json');
+                }
+                if (lang) {
+                    formData.append('language', lang);
+                }
+                const response = await fetch(GROQ_ENDPOINT, {
+                    method: 'POST',
+                    headers: {
+                        Authorization: `Bearer ${apiKey}`,
+                    },
+                    body: formData,
+                });
+                if (!response.ok) {
+                    const errorText = await response.text();
+                    onError(`Groq Whisper API error: ${response.status} ${errorText}`);
+                    return;
+                }
+                const data = (await response.json());
+                const transcript = {
+                    text: data.text,
+                    isFinal: true,
+                    lang: data.language,
+                };
+                // Parse word-level timestamps when requested and available
+                if (options?.requestWords && data.words && data.words.length > 0) {
+                    const words = data.words.map((w) => ({
+                        word: w.word,
+                        startMs: w.start * 1000,
+                        durationMs: (w.end - w.start) * 1000,
+                    }));
+                    transcript.words = words;
+                }
+                onTranscript(transcript);
+            },
+        };
+    },
+};
+//# sourceMappingURL=GroqWhisper.js.map

package/dist/server/audio/stt/GroqWhisper.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"GroqWhisper.js","sourceRoot":"","sources":["../../../../src/server/audio/stt/GroqWhisper.ts"],"names":[],"mappings":"AAQA,MAAM,aAAa,GAAG,qDAAqD,CAAC;AAE5E,SAAS,cAAc,CAAC,UAAkB;IACxC,MAAM,UAAU,GAAG,KAAK,CAAC;IACzB,MAAM,WAAW,GAAG,CAAC,CAAC;IACtB,MAAM,aAAa,GAAG,EAAE,CAAC;IACzB,MAAM,QAAQ,GAAG,UAAU,GAAG,WAAW,GAAG,CAAC,aAAa,GAAG,CAAC,CAAC,CAAC;IAChE,MAAM,UAAU,GAAG,WAAW,GAAG,CAAC,aAAa,GAAG,CAAC,CAAC,CAAC;IACrD,MAAM,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC;IAEhC,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IACxB,MAAM,CAAC,aAAa,CAAC,EAAE,GAAG,UAAU,EAAE,CAAC,CAAC,CAAC;IACzC,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;IACxB,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;IACzB,MAAM,CAAC,aAAa,CAAC,EAAE,EAAE,EAAE,CAAC,CAAC,CAAC,iBAAiB;IAC/C,MAAM,CAAC,aAAa,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,aAAa;IAC1C,MAAM,CAAC,aAAa,CAAC,WAAW,EAAE,EAAE,CAAC,CAAC;IACtC,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;IACrC,MAAM,CAAC,aAAa,CAAC,QAAQ,EAAE,EAAE,CAAC,CAAC;IACnC,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;IACrC,MAAM,CAAC,aAAa,CAAC,aAAa,EAAE,EAAE,CAAC,CAAC;IACxC,MAAM,CAAC,KAAK,CAAC,MAAM,EAAE,EAAE,CAAC,CAAC;IACzB,MAAM,CAAC,aAAa,CAAC,UAAU,EAAE,EAAE,CAAC,CAAC;IAErC,OAAO,MAAM,CAAC;AAChB,CAAC;AAcD,MAAM,CAAC,MAAM,sBAAsB,GAAgB;IACjD,IAAI,EAAE,cAAc;IACpB,SAAS,EAAE,cAAc;IAEzB,KAAK,CAAC,OAAO,CACX,YAA6C,EAC7C,OAA8B,EAC9B,IAAa,EACb,OAA2B;QAE3B,MAAM,MAAM,GAAG,OAAO,CAAC,GAAG,CAAC,cAAc,CAAC,CAAC;QAC3C,IAAI,CAAC,MAAM,EAAE,CAAC;YACZ,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAC;QAC7C,CAAC;QAED,MAAM,MAAM,GAAa,EAAE,CAAC;QAE5B,OAAO;YACL,SAAS,CAAC,KAAa;gBACrB,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC;YACrB,CAAC;YAED,KAAK,CAAC,IAAI;gBACR,IAAI,MAAM,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;oBACxB,OAAO;gBACT,CAAC;gBAED,MAAM,OAAO,GAAG,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC;gBACtC,MAAM,SAAS,GAAG,cAAc,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;gBACjD,MAAM,OAAO,GAAG,MAAM,CAAC,MAAM,CAAC,CAAC,SAAS,EAAE,OAAO,CAAC,CAAC,CAAC;gBAEpD,MAAM,QAAQ,GAAG,IAAI,QAAQ,EAAE,CAAC;gBAChC,QAAQ,CAAC,MAAM,CACb,MAAM,EACN,IAAI,IAAI,CAAC,CAAC,OAAO,CAAC,EAAE,EAAE,IAAI,EAAE,WAAW,EAAE,CAAC,EAC1C,WAAW,CACZ,CAAC;gBACF,QAAQ,CAAC,MAAM,CAAC,OAAO,EAAE,wBAAwB,CAAC,CAAC;gBAEnD,yFAAyF;gBACzF,iDAAiD;gBACjD,IAAI,OAAO,EAAE,YAAY,EAAE,CAAC;oBAC1B,QAAQ,CAAC,MAAM,CAAC,iBAAiB,EAAE,cAAc,CAAC,CAAC;oBACnD,QAAQ,CAAC,MAAM,CAAC,2BAA2B,EAAE,MAAM,CAAC,CAAC;gBACvD,CAAC;qBAAM,CAAC;oBACN,QAAQ,CAAC,MAAM,CAAC,iBAAiB,EAAE,MAAM,CAAC,CAAC;gBAC7C,CAAC;gBAED,IAAI,IAAI,EAAE,CAAC;oBACT,QAAQ,CAAC,MAAM,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC;gBACpC,CAAC;gBAED,MAAM,QAAQ,GAAG,MAAM,KAAK,CAAC,aAAa,EAAE;oBAC1C,MAAM,EAAE,MAAM;oBACd,OAAO,EAAE;wBACP,aAAa,EAAE,UAAU,MAAM,EAAE;qBAClC;oBACD,IAAI,EAAE,QAAQ;iBACf,CAAC,CAAC;gBAEH,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;oBACjB,MAAM,SAAS,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAC;oBACxC,OAAO,CAAC,2BAA2B,QAAQ,CAAC,MAAM,IAAI,SAAS,EAAE,CAAC,CAAC;oBACnE,OAAO;gBACT,CAAC;gBAED,MAAM,IAAI,GAAG,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,CAAwB,CAAC;gBAE5D,MAAM,UAAU,GAAkB;oBAChC,IAAI,EAAE,IAAI,CAAC,IAAI;oBACf,OAAO,EAAE,IAAI;oBACb,IAAI,EAAE,IAAI,CAAC,QAAQ;iBACpB,CAAC;gBAEF,2DAA2D;gBAC3D,IAAI,OAAO,EAAE,YAAY,IAAI,IAAI,CAAC,KAAK,IAAI,IAAI,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;oBACjE,MAAM,KAAK,GAAc,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;wBAC9C,IAAI,EAAE,CAAC,CAAC,IAAI;wBACZ,OAAO,EAAE,CAAC,CAAC,KAAK,GAAG,IAAI;wBACvB,UAAU,EAAE,CAAC,CAAC,CAAC,GAAG,GAAG,CAAC,CAAC,KAAK,CAAC,GAAG,IAAI;qBACrC,CAAC,CAAC,CAAC;oBACJ,UAAU,CAAC,KAAK,GAAG,KAAK,CAAC;gBAC3B,CAAC;gBAED,YAAY,CAAC,UAAU,CAAC,CAAC;YAC3B,CAAC;SACF,CAAC;IACJ,CAAC;CACF,CAAC"}

package/dist/server/audio/stt/registry.d.ts

@@ -0,0 +1,5 @@
+import type { STTProvider } from './types.js';
+export declare function registerSTTProvider(provider: STTProvider): void;
+export declare function selectSTTProvider(): STTProvider | undefined;
+export declare function getAllSTTProviders(): STTProvider[];
+//# sourceMappingURL=registry.d.ts.map

package/dist/server/audio/stt/registry.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"registry.d.ts","sourceRoot":"","sources":["../../../../src/server/audio/stt/registry.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,YAAY,CAAC;AAI9C,wBAAgB,mBAAmB,CAAC,QAAQ,EAAE,WAAW,GAAG,IAAI,CAE/D;AAED,wBAAgB,iBAAiB,IAAI,WAAW,GAAG,SAAS,CAE3D;AAED,wBAAgB,kBAAkB,IAAI,WAAW,EAAE,CAElD"}

package/dist/server/audio/stt/registry.js

@@ -0,0 +1,11 @@
+const providers = [];
+export function registerSTTProvider(provider) {
+    providers.unshift(provider);
+}
+export function selectSTTProvider() {
+    return providers.find((p) => !!process.env[p.apiKeyEnv]);
+}
+export function getAllSTTProviders() {
+    return [...providers];
+}
+//# sourceMappingURL=registry.js.map

package/dist/server/audio/stt/registry.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"registry.js","sourceRoot":"","sources":["../../../../src/server/audio/stt/registry.ts"],"names":[],"mappings":"AAEA,MAAM,SAAS,GAAkB,EAAE,CAAC;AAEpC,MAAM,UAAU,mBAAmB,CAAC,QAAqB;IACvD,SAAS,CAAC,OAAO,CAAC,QAAQ,CAAC,CAAC;AAC9B,CAAC;AAED,MAAM,UAAU,iBAAiB;IAC/B,OAAO,SAAS,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,CAAC,CAAC,SAAS,CAAC,CAAC,CAAC;AAC3D,CAAC;AAED,MAAM,UAAU,kBAAkB;IAChC,OAAO,CAAC,GAAG,SAAS,CAAC,CAAC;AACxB,CAAC"}

package/dist/server/audio/stt/types.d.ts

@@ -0,0 +1,25 @@
+import type { STTWord } from '../../../shared/Audio.js';
+export interface STTTranscript {
+    text: string;
+    isFinal: boolean;
+    lang?: string;
+    /** Word-level timestamps — only present when requestWords was true */
+    words?: STTWord[];
+}
+export interface STTSession {
+    /** Send raw PCM16 audio at 16kHz mono (Buffer) */
+    sendAudio(pcm16: Buffer): void;
+    /** Flush and get final transcript */
+    stop(): Promise<void>;
+}
+export interface STTConnectOptions {
+    /** Request word-level timestamps. Default false — Groq verbose_json is slower */
+    requestWords?: boolean;
+}
+export interface STTProvider {
+    name: string;
+    apiKeyEnv: string;
+    /** Open a session; onTranscript is called with each result */
+    connect(onTranscript: (result: STTTranscript) => void, onError: (err: string) => void, lang?: string, options?: STTConnectOptions): Promise<STTSession>;
+}
+//# sourceMappingURL=types.d.ts.map

package/dist/server/audio/stt/types.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../../../../src/server/audio/stt/types.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,OAAO,EAAE,MAAM,0BAA0B,CAAC;AAExD,MAAM,WAAW,aAAa;IAC5B,IAAI,EAAE,MAAM,CAAC;IACb,OAAO,EAAE,OAAO,CAAC;IACjB,IAAI,CAAC,EAAE,MAAM,CAAC;IACd,sEAAsE;IACtE,KAAK,CAAC,EAAE,OAAO,EAAE,CAAC;CACnB;AAED,MAAM,WAAW,UAAU;IACzB,kDAAkD;IAClD,SAAS,CAAC,KAAK,EAAE,MAAM,GAAG,IAAI,CAAC;IAC/B,qCAAqC;IACrC,IAAI,IAAI,OAAO,CAAC,IAAI,CAAC,CAAC;CACvB;AAED,MAAM,WAAW,iBAAiB;IAChC,iFAAiF;IACjF,YAAY,CAAC,EAAE,OAAO,CAAC;CACxB;AAED,MAAM,WAAW,WAAW;IAC1B,IAAI,EAAE,MAAM,CAAC;IACb,SAAS,EAAE,MAAM,CAAC;IAClB,8DAA8D;IAC9D,OAAO,CACL,YAAY,EAAE,CAAC,MAAM,EAAE,aAAa,KAAK,IAAI,EAC7C,OAAO,EAAE,CAAC,GAAG,EAAE,MAAM,KAAK,IAAI,EAC9B,IAAI,CAAC,EAAE,MAAM,EACb,OAAO,CAAC,EAAE,iBAAiB,GAC1B,OAAO,CAAC,UAAU,CAAC,CAAC;CACxB"}

package/dist/server/audio/stt/types.js

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

package/dist/server/audio/stt/types.js.map

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

package/dist/server/audio/stt/vad.d.ts

@@ -0,0 +1,18 @@
+export interface VADProcessor {
+    isSpeech(pcm16Buffer: Buffer): Promise<boolean>;
+    reset(): void;
+}
+/**
+ * Load the Silero VAD ONNX model from the given path and return a stateful VADProcessor.
+ * One VADProcessor per STT session — the LSTM state is maintained across calls.
+ *
+ * The Silero VAD v5 model expects:
+ *   - input:  float32 PCM at 16kHz, shape [1, numSamples] (with context prepended)
+ *   - state:  float32 combined LSTM state, shape [2, 1, 128]
+ *   - sr:     int64 sample rate, value 16000
+ * Outputs:
+ *   - output:  float32 speech probability scalar
+ *   - stateN:  updated LSTM state, shape [2, 1, 128]
+ */
+export declare function loadVADModel(modelPath: string): Promise<VADProcessor>;
+//# sourceMappingURL=vad.d.ts.map

package/dist/server/audio/stt/vad.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"vad.d.ts","sourceRoot":"","sources":["../../../../src/server/audio/stt/vad.ts"],"names":[],"mappings":"AAEA,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,WAAW,EAAE,MAAM,GAAG,OAAO,CAAC,OAAO,CAAC,CAAC;IAChD,KAAK,IAAI,IAAI,CAAC;CACf;AAED;;;;;;;;;;;GAWG;AACH,wBAAsB,YAAY,CAAC,SAAS,EAAE,MAAM,GAAG,OAAO,CAAC,YAAY,CAAC,CAO3E"}

package/dist/server/audio/stt/vad.js

@@ -0,0 +1,90 @@
+import * as ort from 'onnxruntime-node';
+/**
+ * Load the Silero VAD ONNX model from the given path and return a stateful VADProcessor.
+ * One VADProcessor per STT session — the LSTM state is maintained across calls.
+ *
+ * The Silero VAD v5 model expects:
+ *   - input:  float32 PCM at 16kHz, shape [1, numSamples] (with context prepended)
+ *   - state:  float32 combined LSTM state, shape [2, 1, 128]
+ *   - sr:     int64 sample rate, value 16000
+ * Outputs:
+ *   - output:  float32 speech probability scalar
+ *   - stateN:  updated LSTM state, shape [2, 1, 128]
+ */
+export async function loadVADModel(modelPath) {
+    const session = await ort.InferenceSession.create(modelPath, {
+        executionProviders: ['cpu'],
+        graphOptimizationLevel: 'all',
+    });
+    return new SileroVADProcessor(session);
+}
+const SAMPLE_RATE = 16000;
+const CHUNK_SIZE = 512; // 32 ms at 16 kHz
+const CONTEXT_SIZE = 64; // temporal context required by Silero VAD at 16 kHz
+// Silero VAD v5 combined LSTM state: shape [2, 1, 128]
+const STATE_SIZE = 256; // 2 * 1 * 128
+class SileroVADProcessor {
+    session;
+    state;
+    context;
+    constructor(session) {
+        this.session = session;
+        this.state = new ort.Tensor('float32', new Float32Array(STATE_SIZE), [2, 1, 128]);
+        this.context = new Float32Array(CONTEXT_SIZE);
+    }
+    /**
+     * Run VAD on a PCM16 buffer.
+     * The buffer may contain any number of samples; if it doesn't align to
+     * CHUNK_SIZE (512 samples = 1024 bytes) the last partial chunk is zero-padded.
+     * Returns true if speech is detected in any chunk of the buffer.
+     */
+    async isSpeech(pcm16Buffer) {
+        const numSamples = Math.floor(pcm16Buffer.length / 2);
+        if (numSamples === 0)
+            return false;
+        // Convert PCM16 → float32 in [-1, 1]
+        const float32 = new Float32Array(numSamples);
+        for (let i = 0; i < numSamples; i++) {
+            float32[i] = pcm16Buffer.readInt16LE(i * 2) / 32768;
+        }
+        // Process in CHUNK_SIZE windows; zero-pad the last chunk if needed
+        let speechDetected = false;
+        const numChunks = Math.ceil(numSamples / CHUNK_SIZE);
+        for (let c = 0; c < numChunks; c++) {
+            const chunk = new Float32Array(CHUNK_SIZE);
+            const srcOffset = c * CHUNK_SIZE;
+            const available = Math.min(CHUNK_SIZE, numSamples - srcOffset);
+            chunk.set(float32.subarray(srcOffset, srcOffset + available));
+            // remaining bytes stay zero-padded
+            // Prepend context to form the model input
+            const inputData = new Float32Array(CONTEXT_SIZE + CHUNK_SIZE);
+            inputData.set(this.context, 0);
+            inputData.set(chunk, CONTEXT_SIZE);
+            const inputTensor = new ort.Tensor('float32', inputData, [
+                1,
+                inputData.length,
+            ]);
+            const srTensor = new ort.Tensor('int64', BigInt64Array.from([BigInt(SAMPLE_RATE)]), []);
+            const results = await this.session.run({
+                input: inputTensor,
+                state: this.state,
+                sr: srTensor,
+            });
+            // Update recurrent state for next chunk
+            this.state = results['stateN'];
+            // Slide context window forward
+            this.context = inputData.slice(-CONTEXT_SIZE);
+            const prob = results['output'].data[0];
+            if (prob > 0.5) {
+                speechDetected = true;
+            }
+        }
+        return speechDetected;
+    }
+    /** Zero the LSTM state and context — call between utterances if needed */
+    reset() {
+        this.state = new ort.Tensor('float32', new Float32Array(STATE_SIZE), [2, 1, 128]);
+        this.context = new Float32Array(CONTEXT_SIZE);
+    }
+}
+//# sourceMappingURL=vad.js.map

package/dist/server/audio/stt/vad.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"vad.js","sourceRoot":"","sources":["../../../../src/server/audio/stt/vad.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,GAAG,MAAM,kBAAkB,CAAC;AAOxC;;;;;;;;;;;GAWG;AACH,MAAM,CAAC,KAAK,UAAU,YAAY,CAAC,SAAiB;IAClD,MAAM,OAAO,GAAG,MAAM,GAAG,CAAC,gBAAgB,CAAC,MAAM,CAAC,SAAS,EAAE;QAC3D,kBAAkB,EAAE,CAAC,KAAK,CAAC;QAC3B,sBAAsB,EAAE,KAAK;KAC9B,CAAC,CAAC;IAEH,OAAO,IAAI,kBAAkB,CAAC,OAAO,CAAC,CAAC;AACzC,CAAC;AAED,MAAM,WAAW,GAAG,KAAK,CAAC;AAC1B,MAAM,UAAU,GAAG,GAAG,CAAC,CAAC,kBAAkB;AAC1C,MAAM,YAAY,GAAG,EAAE,CAAC,CAAC,oDAAoD;AAC7E,uDAAuD;AACvD,MAAM,UAAU,GAAG,GAAG,CAAC,CAAC,cAAc;AAEtC,MAAM,kBAAkB;IACL,OAAO,CAAuB;IACvC,KAAK,CAAa;IAClB,OAAO,CAAe;IAE9B,YAAY,OAA6B;QACvC,IAAI,CAAC,OAAO,GAAG,OAAO,CAAC;QACvB,IAAI,CAAC,KAAK,GAAG,IAAI,GAAG,CAAC,MAAM,CACzB,SAAS,EACT,IAAI,YAAY,CAAC,UAAU,CAAC,EAC5B,CAAC,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,CACZ,CAAC;QACF,IAAI,CAAC,OAAO,GAAG,IAAI,YAAY,CAAC,YAAY,CAAC,CAAC;IAChD,CAAC;IAED;;;;;OAKG;IACH,KAAK,CAAC,QAAQ,CAAC,WAAmB;QAChC,MAAM,UAAU,GAAG,IAAI,CAAC,KAAK,CAAC,WAAW,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;QACtD,IAAI,UAAU,KAAK,CAAC;YAAE,OAAO,KAAK,CAAC;QAEnC,qCAAqC;QACrC,MAAM,OAAO,GAAG,IAAI,YAAY,CAAC,UAAU,CAAC,CAAC;QAC7C,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,UAAU,EAAE,CAAC,EAAE,EAAE,CAAC;YACpC,OAAO,CAAC,CAAC,CAAC,GAAG,WAAW,CAAC,WAAW,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,KAAK,CAAC;QACtD,CAAC;QAED,mEAAmE;QACnE,IAAI,cAAc,GAAG,KAAK,CAAC;QAC3B,MAAM,SAAS,GAAG,IAAI,CAAC,IAAI,CAAC,UAAU,GAAG,UAAU,CAAC,CAAC;QAErD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,SAAS,EAAE,CAAC,EAAE,EAAE,CAAC;YACnC,MAAM,KAAK,GAAG,IAAI,YAAY,CAAC,UAAU,CAAC,CAAC;YAC3C,MAAM,SAAS,GAAG,CAAC,GAAG,UAAU,CAAC;YACjC,MAAM,SAAS,GAAG,IAAI,CAAC,GAAG,CAAC,UAAU,EAAE,UAAU,GAAG,SAAS,CAAC,CAAC;YAC/D,KAAK,CAAC,GAAG,CAAC,OAAO,CAAC,QAAQ,CAAC,SAAS,EAAE,SAAS,GAAG,SAAS,CAAC,CAAC,CAAC;YAC9D,mCAAmC;YAEnC,0CAA0C;YAC1C,MAAM,SAAS,GAAG,IAAI,YAAY,CAAC,YAAY,GAAG,UAAU,CAAC,CAAC;YAC9D,SAAS,CAAC,GAAG,CAAC,IAAI,CAAC,OAAO,EAAE,CAAC,CAAC,CAAC;YAC/B,SAAS,CAAC,GAAG,CAAC,KAAK,EAAE,YAAY,CAAC,CAAC;YAEnC,MAAM,WAAW,GAAG,IAAI,GAAG,CAAC,MAAM,CAAC,SAAS,EAAE,SAAS,EAAE;gBACvD,CAAC;gBACD,SAAS,CAAC,MAAM;aACjB,CAAC,CAAC;YACH,MAAM,QAAQ,GAAG,IAAI,GAAG,CAAC,MAAM,CAC7B,OAAO,EACP,aAAa,CAAC,IAAI,CAAC,CAAC,MAAM,CAAC,WAAW,CAAC,CAAC,CAAC,EACzC,EAAE,CACH,CAAC;YAEF,MAAM,OAAO,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,GAAG,CAAC;gBACrC,KAAK,EAAE,WAAW;gBAClB,KAAK,EAAE,IAAI,CAAC,KAAK;gBACjB,EAAE,EAAE,QAAQ;aACb,CAAC,CAAC;YAEH,wCAAwC;YACxC,IAAI,CAAC,KAAK,GAAG,OAAO,CAAC,QAAQ,CAAE,CAAC;YAChC,+BAA+B;YAC/B,IAAI,CAAC,OAAO,GAAG,SAAS,CAAC,KAAK,CAAC,CAAC,YAAY,CAAC,CAAC;YAE9C,MAAM,IAAI,GAAG,OAAO,CAAC,QAAQ,CAAE,CAAC,IAAI,CAAC,CAAC,CAAW,CAAC;YAClD,IAAI,IAAI,GAAG,GAAG,EAAE,CAAC;gBACf,cAAc,GAAG,IAAI,CAAC;YACxB,CAAC;QACH,CAAC;QAED,OAAO,cAAc,CAAC;IACxB,CAAC;IAED,0EAA0E;IAC1E,KAAK;QACH,IAAI,CAAC,KAAK,GAAG,IAAI,GAAG,CAAC,MAAM,CACzB,SAAS,EACT,IAAI,YAAY,CAAC,UAAU,CAAC,EAC5B,CAAC,CAAC,EAAE,CAAC,EAAE,GAAG,CAAC,CACZ,CAAC;QACF,IAAI,CAAC,OAAO,GAAG,IAAI,YAAY,CAAC,YAAY,CAAC,CAAC;IAChD,CAAC;CACF"}

package/dist/server/audio/tts/InWorld.d.ts

@@ -0,0 +1,3 @@
+import type { TTSProvider } from './types.js';
+export declare const inworldTTSProvider: TTSProvider;
+//# sourceMappingURL=InWorld.d.ts.map

package/dist/server/audio/tts/InWorld.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"InWorld.d.ts","sourceRoot":"","sources":["../../../../src/server/audio/tts/InWorld.ts"],"names":[],"mappings":"AAEA,OAAO,KAAK,EAAY,WAAW,EAAoB,MAAM,YAAY,CAAC;AAuE1E,eAAO,MAAM,kBAAkB,EAAE,WAsHhC,CAAC"}