0cb1cac7ae
Issue: transformers.js is an ES module package and cannot be loaded with importScripts() in classic workers. Solution: - Changed to module worker (type: 'module') - Import transformers.js from CDN as ES module - Models (42MB) still served from local server at /models/ Trade-off: - Library (900KB): Loads from cdn.jsdelivr.net once, cached - Models (42MB): Self-hosted, served from /models/ (no CDN) This is necessary because: 1. @xenova/transformers is ES module-only (package.json: "type": "module") 2. ES modules cannot use importScripts() 3. Module workers require HTTPS for imports 4. CDN is HTTPS and cacheable If CDN is blocked: - Use Web Speech API (with privacy warnings) - OR use Server Transcription (Vertex AI/AWS) Models remain self-hosted as they're 40MB+ and contain the AI.
84 lines
3 KiB
JavaScript
84 lines
3 KiB
JavaScript
// ============================================================
|
|
// WHISPER WORKER — runs @xenova/transformers in a Web Worker
|
|
// Audio never leaves the device. Model cached in IndexedDB.
|
|
// Uses ES module worker (type: 'module')
|
|
// ============================================================
|
|
|
|
import { pipeline, env } from 'https://cdn.jsdelivr.net/npm/@xenova/transformers@2.17.2';
|
|
|
|
console.log('[WhisperWorker] Starting worker initialization');
|
|
|
|
// Configure transformers.js to load models from our local server
|
|
env.allowLocalModels = false;
|
|
env.useBrowserCache = true;
|
|
env.allowRemoteModels = false; // Force local-only
|
|
env.localModelPath = '/models/'; // Our bundled models (40MB+)
|
|
console.log('[WhisperWorker] Transformers configured for local models at /models/');
|
|
|
|
let _pipe = null;
|
|
let _loadedModel = null;
|
|
|
|
async function load(modelName) {
|
|
if (_pipe && _loadedModel === modelName) {
|
|
console.log('[WhisperWorker] Model already loaded:', modelName);
|
|
return;
|
|
}
|
|
console.log('[WhisperWorker] Loading model:', modelName);
|
|
self.postMessage({ type: 'loading', model: modelName });
|
|
|
|
try {
|
|
_pipe = await pipeline('automatic-speech-recognition', modelName, {
|
|
progress_callback: (p) => {
|
|
console.log('[WhisperWorker] Progress:', p.status, p.file, p.progress);
|
|
if (p.status === 'downloading' || p.status === 'progress') {
|
|
self.postMessage({ type: 'progress', file: p.file || '', progress: Math.round(p.progress || 0) });
|
|
}
|
|
if (p.status === 'done' || p.status === 'ready') {
|
|
self.postMessage({ type: 'progress', file: p.file || '', progress: 100 });
|
|
}
|
|
}
|
|
});
|
|
_loadedModel = modelName;
|
|
console.log('[WhisperWorker] Model loaded successfully');
|
|
self.postMessage({ type: 'ready', model: modelName });
|
|
} catch (err) {
|
|
console.error('[WhisperWorker] Load error:', err);
|
|
self.postMessage({ type: 'error', message: 'Model load failed: ' + err.message });
|
|
}
|
|
}
|
|
|
|
self.addEventListener('message', async (e) => {
|
|
const d = e.data;
|
|
console.log('[WhisperWorker] Message received:', d.type);
|
|
|
|
if (d.type === 'load') {
|
|
try {
|
|
await load(d.model || 'Xenova/whisper-tiny.en');
|
|
} catch(err) {
|
|
console.error('[WhisperWorker] Load error:', err);
|
|
self.postMessage({ type: 'error', message: 'Load failed: ' + err.message });
|
|
}
|
|
return;
|
|
}
|
|
|
|
if (d.type === 'transcribe') {
|
|
try {
|
|
await load(d.model || 'Xenova/whisper-tiny.en');
|
|
console.log('[WhisperWorker] Starting transcription...');
|
|
const result = await _pipe(d.audio, {
|
|
language: 'english',
|
|
task: 'transcribe',
|
|
chunk_length_s: 30,
|
|
stride_length_s: 5,
|
|
return_timestamps: false
|
|
});
|
|
console.log('[WhisperWorker] Transcription complete');
|
|
self.postMessage({ type: 'result', text: result.text.trim() });
|
|
} catch(err) {
|
|
console.error('[WhisperWorker] Transcribe error:', err);
|
|
self.postMessage({ type: 'error', message: 'Transcription failed: ' + err.message });
|
|
}
|
|
}
|
|
});
|
|
|
|
console.log('[WhisperWorker] Worker initialized');
|