snxraven/gnome-jarvis
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 1 Wiki Activity

Check Point
Rolling release / release (push) Successful in 6m36s
Details

Browse Source
This commit is contained in:
snxraven
2026-09-12 09:07:09 -04:00
parent e9040d110a
commit a4073b9020
63 changed files with 2459 additions and 632 deletions
Download Patch File Download Diff File Expand all files Collapse all files
daemon/voice-loop.js
+65 -16
View File
@@ -20,18 +20,23 @@ const FAST_COMMANDS = new Map([
export function fastCommand(text) { return FAST_COMMANDS.get(String(text || '').trim().toLowerCase()) || null; }
export class VoiceLoop extends EventEmitter {
constructor({ daemon, capture = new PipeWireCapture(), playback = new PipeWirePlayback(), wake = new WakeEngine(), vad = new VadSegmenter(), asr, tts, cooldownMs = POST_PLAYBACK_COOLDOWN_MS, now = () => Date.now() } = {}) {
super(); this.daemon = daemon; this.capture = capture; this.playback = playback; this.wake = wake; this.vad = vad; this.asr = asr; this.tts = tts; this.cooldownMs = cooldownMs; this.now = now;
this.isSpeaking = false; this.cooldownUntil = 0; this.running = false; this.ptt = false; this._generation = 0; this._speechQueue = Promise.resolve(); this.metrics = new VoiceMetrics();
constructor({ daemon, capture = new PipeWireCapture(), playback = new PipeWirePlayback(), wake = new WakeEngine(), vad = new VadSegmenter(), asr, tts, cooldownMs = POST_PLAYBACK_COOLDOWN_MS, listeningMode = 'conversation', now = () => Date.now(), captureRetryMs = 2000, asrRetryMs = 4000 } = {}) {
super(); this.daemon = daemon; this.capture = capture; this.playback = playback; this.wake = wake; this.vad = vad; this.asr = asr; this.tts = tts; this.cooldownMs = cooldownMs; this.listeningMode = listeningMode; this.now = now;
this.captureRetryMs = captureRetryMs; this.asrRetryMs = asrRetryMs;
this.isSpeaking = false; this.cooldownUntil = 0; this.running = false; this.ptt = false; this._generation = 0; this._speechQueue = Promise.resolve(); this.metrics = new VoiceMetrics(); this._transcriptions = new Set();
capture.on('audio', (chunk) => this.pushAudio(chunk));
capture.on('error', (error) => { this.status.capture = false; this.status.errors.capture = error.message; this.emit('error', error); });
capture.on('close', () => { this.status.capture = false; if (this.running) { this.status.errors.capture = 'Microphone stream closed'; this.emit('error', new Error('Microphone stream closed')); } });
capture.on('error', (error) => { this.status.capture = false; this.status.errors.capture = error.message; this.emit('error', error); this._scheduleCaptureRetry(); });
capture.on('close', () => { this.status.capture = false; if (this.running) { this.status.errors.capture = 'Microphone stream closed'; this._scheduleCaptureRetry(); } });
wake.on('unavailable', () => { this.status.wake = false; });
wake.on('error', (error) => { this.status.wake = false; this.emit('error', error); });
this.status = { asr: false, tts: false, capture: false, wake: false, errors: {} };
wake.on('wake', (phrase) => this.wakeHeard(phrase));
vad.on('level', (rms) => daemon?.emit('ListeningLevel', rms));
vad.on('utterance', (audio) => this.transcribe(audio).catch((error) => this.emit('error', error)));
vad.on('utterance', (audio) => {
const task = this.transcribe(audio).catch((error) => this.emit('error', error));
this._transcriptions.add(task);
task.finally(() => this._transcriptions.delete(task)).catch(() => {});
});
}
async start() {
@@ -48,20 +53,62 @@ export class VoiceLoop extends EventEmitter {
}
this.running = true;
if (this.status.asr) {
try { this.capture.start(); this.status.capture = true; }
catch (error) { this.status.errors.capture = error.message; this.emit('error', error); }
try { this.wake.start?.(); this.wake.resume(); this.status.wake = Boolean(this.wake.command || this.wake.detect); }
catch (error) { this.status.errors.wake = error.message; this.emit('error', error); }
this._armCapture();
this._armWake();
} else if (this.asr) {
this._scheduleAsrRetry();
}
}
async stop() { this.running = false; this.capture.stop(); this.wake.close(); this.vad.reset(); this.playback.stop(); await this.asr?.stop?.(); if (this.tts !== this.asr) await this.tts?.stop?.(); }
_notifyStatus() { try { this.daemon?.emit?.('StateChanged', this.daemon.state || 'ARMED'); } catch {} }
_armWake() {
try { if (this.listeningMode !== 'ptt') this.wake.start?.(); this.wake.resume(); this.status.wake = this.listeningMode !== 'ptt' && Boolean(this.wake.command || this.wake.detect); }
catch (error) { this.status.errors.wake = error.message; this.emit('error', error); }
}
_armCapture() {
if (!this.running || !this.status.asr || this.status.capture) return;
try { this.capture.start(); this.status.capture = true; delete this.status.errors.capture; this._notifyStatus(); }
catch (error) { this.status.errors.capture = error.message; this.emit('error', error); this._scheduleCaptureRetry(); }
}
_scheduleCaptureRetry() {
if (!this.running || this._captureRetry || this.status.capture || !this.status.asr) return;
this._captureRetry = setTimeout(() => {
this._captureRetry = 0;
this._armCapture();
}, this.captureRetryMs);
this._captureRetry.unref?.();
}
_scheduleAsrRetry() {
if (!this.running || this._asrRetry || this.status.asr || !this.asr) return;
this._asrRetry = setTimeout(() => {
this._asrRetry = 0;
if (!this.running || this.status.asr || !this.asr) return;
this.asr.start().then(() => {
if (!this.running || this.status.asr) return;
this.status.asr = true;
delete this.status.errors.asr;
this._armCapture();
this._armWake();
this._notifyStatus();
}).catch((error) => {
this.status.errors.asr = String(error?.message || error);
this._scheduleAsrRetry();
});
}, this.asrRetryMs);
this._asrRetry.unref?.();
}
async stop() {
this.running = false;
if (this._captureRetry) { clearTimeout(this._captureRetry); this._captureRetry = 0; }
if (this._asrRetry) { clearTimeout(this._asrRetry); this._asrRetry = 0; }
this.interrupt(); this.capture.stop(); this.wake.close(); this.vad.reset(); this.playback.stop(); await Promise.allSettled([this._speechQueue, ...this._transcriptions]); await this.asr?.stop?.(); if (this.tts !== this.asr) await this.tts?.stop?.();
}
setPushToTalk(pressed) { this.ptt = Boolean(pressed) && this.status.asr && this.status.capture; if (this.ptt) this.wakeHeard('push-to-talk'); else this.vad.end(); }
pushAudio(chunk) {
if (!this.running || this.daemon?.locked || this.daemon?.state === 'SLEEPING') return;
if (this.isSpeaking || this.now() < this.cooldownUntil) { this.metrics.feedbackDrop(); return; }
try { this.daemon?.emit('ListeningLevel', Math.min(1, Math.max(0, chunk?.length ? 0.01 : 0))); } catch {}
if (!this.ptt) this.wake.push(chunk);
if (this.ptt || this.daemon?.state === 'LISTENING') this.vad.push(chunk);
if (!this.ptt && this.listeningMode !== 'ptt') this.wake.push(chunk);
if (this.ptt || (this.listeningMode !== 'ptt' && this.daemon?.state === 'LISTENING')) this.vad.push(chunk);
}
wakeHeard(phrase) {
if (!this.running || this.daemon?.locked) return;
@@ -69,8 +116,9 @@ export class VoiceLoop extends EventEmitter {
}
async transcribe(audio) {
if (!this.status.asr || !this.asr?.transcribeAudio) return;
const generation = this._generation;
const text = await this.asr.transcribeAudio(audio).catch((error) => { this.emit('error', error); return ''; });
if (!this.running || this.daemon?.locked) return;
if (!this.running || this.daemon?.locked || generation !== this._generation) return;
if (!isMeaningfulTranscript(text)) { this.metrics.wakeRejected(); return; }
this.metrics.utterance();
this.daemon?.emit('PartialTranscript', text); this.daemon?.emit('FinalTranscript', text);
@@ -93,7 +141,8 @@ export class VoiceLoop extends EventEmitter {
this.wake.resume();
if (this.daemon?.state === 'SPEAKING') {
try { this.daemon.voice?.finishSpeaking?.(); } catch {}
this.daemon.setState?.('LISTENING');
if (this.daemon._finishSpeech) this.daemon._finishSpeech();
else this.daemon.setState?.('LISTENING');
}
}
async speak(text) {
@@ -127,7 +176,7 @@ export class VoiceLoop extends EventEmitter {
try {
const audio = await this.tts.speak(speakableForTts(text));
if (generation !== this._generation) return;
await this.playback.play(audio.samples);
await this.playback.play(audio.samples, audio.sampleRate);
try { this.daemon?.emit('SpeakingLevel', 0); } catch {}
} finally {
this.isSpeaking = false;