fix: downsample audio to 16kHz to prevent ASR 500 error, add permission check handler, and render errors in compact view (v1.0.8)

This commit is contained in:
Michael committed 2026-06-08 20:44:31 +02:00
1 parent 221bb7062c
commit 5217eee53a
6 files changed
+112 -13

No files matched your search

+2 -2
View File
@@ -1,7 +1,7 @@
# 🚀 EveFlow — Compagnon de Bureau Rétro-Futuriste 3D
[![Windows Version](https://img.shields.io/badge/OS-Windows-blue.svg?style=flat-square&logo=windows)](https://github.com/R0m1k3/EveFlow)
[![Version](https://img.shields.io/badge/version-1.0.7-brightgreen.svg?style=flat-square)](https://github.com/R0m1k3/EveFlow/releases/tag/v1.0.7)
[![Version](https://img.shields.io/badge/version-1.0.8-brightgreen.svg?style=flat-square)](https://github.com/R0m1k3/EveFlow/releases/tag/v1.0.8)
[![License](https://img.shields.io/badge/license-MIT-lightgrey.svg?style=flat-square)](LICENSE)
**EveFlow** est un compagnon de bureau Windows immersif haut de gamme, combinant une esthétique cyberpunk rétro-futuriste soignée et des technologies d'intelligence artificielle avancées. Il intègre un assistant virtuel en 3D nommé **Eve**, animé en temps réel avec des expressions émotionnelles dynamiques et synchronisé avec des services de synthèse vocale (TTS/STT) locaux et des architectures multi-agents.
@@ -74,7 +74,7 @@ Pour générer l'exécutable d'installation Windows (`.exe`) autonome :
```bash
npm run dist
```
*L'exécutable d'installation NSIS (`EveFlow Setup 1.0.7.exe`) et la build décompressée seront générés dans le dossier `./out/`.*
*L'exécutable d'installation NSIS (`EveFlow Setup 1.0.8.exe`) et la build décompressée seront générés dans le dossier `./out/`.*
---
+7
View File
@@ -275,6 +275,13 @@ app.whenReady().then(() => {
callback(false);
});
session.defaultSession.setPermissionCheckHandler((webContents, permission, requestingOrigin, details) => {
if (permission === 'media') {
return true;
}
return false;
});
startWebhookServer();
createWindow();
+2 -2
View File
@@ -1,12 +1,12 @@
{
"name": "eveflow",
"version": "1.0.7",
"version": "1.0.8",
"lockfileVersion": 3,
"requires": true,
"packages": {
"": {
"name": "eveflow",
"version": "1.0.7",
"version": "1.0.8",
"dependencies": {
"@react-three/drei": "^9.105.0",
"@react-three/fiber": "^8.16.0",
+1 -1
View File
@@ -1,6 +1,6 @@
{
"name": "eveflow",
"version": "1.0.7",
"version": "1.0.8",
"description": "Retro-futuristic Windows companion with 3D Eve bot, local TTS/STT, and multi-agent connectivity",
"main": "main.js",
"scripts": {
+21
View File
@@ -2543,6 +2543,27 @@ export const App: React.FC = () => {
<div ref={compactMessagesEndRef} />
</div>
{errorMessage && (
<div style={{
backgroundColor: 'rgba(239, 68, 68, 0.12)',
border: '1px solid rgba(239, 68, 68, 0.35)',
borderRadius: '8px',
color: '#ff6b6b',
padding: '6px 10px',
fontSize: '10px',
display: 'flex',
gap: '6px',
alignItems: 'center',
boxShadow: '0 0 10px rgba(239, 68, 68, 0.05)'
}}>
<AlertCircle style={{ width: '13px', height: '13px', flexShrink: 0, color: '#ef4444' }} />
<div style={{ flex: 1, overflow: 'hidden', textOverflow: 'ellipsis', whiteSpace: 'nowrap' }} title={errorMessage}>
<strong style={{ textTransform: 'uppercase', marginRight: '4px' }}>DÉFAILLANCE:</strong>
{errorMessage}
</div>
</div>
)}
{/* Barre de saisie compacte */}
<div style={{ display: 'flex', gap: '6px', alignItems: 'center', width: '100%', paddingBottom: '2px' }}>
<input
+79 -8
View File
@@ -113,6 +113,19 @@ export class AudioService {
}
}
private logInfo(tag: string, message: string, data?: any) {
console.log(`[${tag}] ${message}`, data);
if ((window as any).electronAPI?.writeLog) {
(window as any).electronAPI.writeLog({
ts: new Date().toISOString(),
level: 'INFO',
tag,
message,
data: data ? { message: data.message, stack: data.stack, ...data } : undefined
});
}
}
constructor() {
this.ttsSynth = window.speechSynthesis;
this.initSTT();
@@ -622,7 +635,18 @@ export class AudioService {
const AudioCtx = window.AudioContext || (window as any).webkitAudioContext;
this.audioContext = new AudioCtx();
// Assurer que l'AudioContext est actif (non suspendu)
if (this.audioContext.state === 'suspended') {
this.logWarn("STT", "AudioContext initialisé à l'état suspendu. Tentative de reprise...");
await this.audioContext.resume();
}
this.sampleRate = this.audioContext.sampleRate;
this.logInfo("STT", "AudioContext démarré avec succès", {
state: this.audioContext.state,
sampleRate: this.sampleRate
});
const source = this.audioContext.createMediaStreamSource(stream);
// bufferSize de 2048, 1 canal d'entrée, 1 de sortie
@@ -687,6 +711,12 @@ export class AudioService {
// Exporter le tampon audio en fichier WAV
const wavBlob = this.exportWAV(this.leftChannel, this.recordingLength, this.sampleRate);
this.logInfo("STT", "Arrêt de l'écoute, exportation WAV...", {
recordingLength: this.recordingLength,
sampleRate: this.sampleRate,
wavSize: wavBlob.size
});
this.transcribeAudioWithAPI(wavBlob, lang)
.then(text => {
if (onResultCallback) onResultCallback(text);
@@ -699,14 +729,19 @@ export class AudioService {
// Fonctions utilitaires pour générer le conteneur WAV PCM 16-bit
private exportWAV(channelData: Float32Array[], recordingLength: number, sampleRate: number): Blob {
const buffer = this.mergeBuffers(channelData, recordingLength);
const wavBuffer = new ArrayBuffer(44 + buffer.length * 2);
const originalBuffer = this.mergeBuffers(channelData, recordingLength);
// Rééchantillonner à 16000 Hz pour la compatibilité avec l'API ASR (Whisper / Qwen3-ASR)
const targetSampleRate = 16000;
const resampledBuffer = this.downsampleBuffer(originalBuffer, sampleRate, targetSampleRate);
const wavBuffer = new ArrayBuffer(44 + resampledBuffer.length * 2);
const view = new DataView(wavBuffer);
/* Identifiant RIFF */
this.writeString(view, 0, 'RIFF');
/* Taille du fichier */
view.setUint32(4, 36 + buffer.length * 2, true);
view.setUint32(4, 36 + resampledBuffer.length * 2, true);
/* Type RIFF */
this.writeString(view, 8, 'WAVE');
/* Identifiant format fmt */
@@ -718,9 +753,9 @@ export class AudioService {
/* Nombre de canaux (1 = Mono) */
view.setUint16(22, 1, true);
/* Fréquence d'échantillonnage */
view.setUint32(24, sampleRate, true);
/* Débit d'octets (sampleRate * blockAlign) */
view.setUint32(28, sampleRate * 2, true);
view.setUint32(24, targetSampleRate, true);
/* Débit d'octets (targetSampleRate * blockAlign) */
view.setUint32(28, targetSampleRate * 2, true);
/* Block align (1 canal * 2 octets par échantillon) */
view.setUint16(32, 2, true);
/* Bits par échantillon (16 bits) */
@@ -728,13 +763,49 @@ export class AudioService {
/* Identifiant de données data */
this.writeString(view, 36, 'data');
/* Longueur des données audio */
view.setUint32(40, buffer.length * 2, true);
view.setUint32(40, resampledBuffer.length * 2, true);
this.floatTo16BitPCM(view, 44, buffer);
this.floatTo16BitPCM(view, 44, resampledBuffer);
return new Blob([view], { type: 'audio/wav' });
}
private downsampleBuffer(buffer: Float32Array, originalSampleRate: number, targetSampleRate: number): Float32Array {
if (originalSampleRate === targetSampleRate) {
return buffer;
}
const sampleRateRatio = originalSampleRate / targetSampleRate;
const newLength = Math.round(buffer.length / sampleRateRatio);
const result = new Float32Array(newLength);
let offsetResult = 0;
let offsetBuffer = 0;
while (offsetResult < result.length) {
const nextOffsetBuffer = Math.round((offsetResult + 1) * sampleRateRatio);
// Moyennage simple (box filter) pour éviter l'aliasing
let accum = 0;
let count = 0;
for (let i = offsetBuffer; i < nextOffsetBuffer && i < buffer.length; i++) {
accum += buffer[i];
count++;
}
if (count > 0) {
result[offsetResult] = accum / count;
} else {
result[offsetResult] = buffer[Math.min(offsetBuffer, buffer.length - 1)];
}
offsetResult++;
offsetBuffer = nextOffsetBuffer;
}
return result;
}
private mergeBuffers(channelData: Float32Array[], recordingLength: number): Float32Array {
const result = new Float32Array(recordingLength);
let offset = 0;