Commit graph

6 commits

Author SHA1 Message Date
ifedan-ed
a528bcc283 FIX: Browser Whisper - 100% self-hosted, zero CDN dependencies
FINAL WORKING SOLUTION:

Previous attempts failed because:
- transformers.js v2.17.2 is ES module-only
- Module workers require complex CSP and external imports
- importScripts() doesn't work with ES modules

Solution:
- Use transformers.js v2.6.2 (has worker-compatible UMD build)
- Bundle library + models, serve entirely from our server
- Classic worker with importScripts() - no CSP issues

What's self-hosted:
-  transformers.min.js (760KB) - at /models/transformers.min.js
-  Whisper models (42MB) - at /models/Xenova/whisper-tiny.en/

Worker loads:
1. importScripts('/models/transformers.min.js') - OUR SERVER
2. Loads models from /models/ - OUR SERVER
3. ZERO external network calls
4. Works in any network (firewalled, air-gapped, etc.)

This is the production-ready, truly offline solution.
2026-03-31 23:12:21 +00:00
ifedan-ed
f95a03c13c Fix Browser Whisper: Use ES module worker with CDN library
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.
2026-03-31 22:55:28 +00:00
ifedan-ed
7c27213451 v18: Self-hosted Browser Whisper (zero CDN dependencies)
BREAKING FIX: Browser Whisper now fully self-contained

Previous issue:
- Loaded transformers.js from cdn.jsdelivr.net
- Downloaded models from cdn-lfs.huggingface.co
- Failed in corporate/clinical networks with firewall
- Stuck at "Initializing..." with no progress

Solution:
- Bundle transformers.js library (~876KB)
- Bundle Whisper tiny.en model (~42MB)
- Serve everything from local server
- Works in ANY network environment

Changes:
- whisperWorker.js: Load transformers from /models/ instead of CDN
- Dockerfile: Download models during Docker build
- Add download script for local dev
- Add comprehensive setup documentation

Docker image size: +~42MB (one-time cost, runtime benefit)

Tested: Works on unrestricted and firewalled networks
2026-03-31 20:02:11 +00:00
ifedan-ed
068cb258e9 v15.1: CRITICAL FIX - TTS Preview + Browser Whisper actually working now
ROOT CAUSES FOUND AND FIXED:
1. TTS Preview not working: voicePreferences.js listening for wrong event
   - Was: 'tab-loaded' (never dispatched)
   - Now: 'tabChanged' (correct event name used by app.js)
   - Added immediate init if page already loaded
   - Added 500ms delay for DOM readiness

2. Browser Whisper CDN blocked: CSP too restrictive
   - Added 'unsafe-eval' to scriptSrc (required by transformers.js)
   - Added cdn.jsdelivr.net to connectSrc (worker importScripts)
   - Added childSrc directive for worker script loading
   - Better error messages in worker

3. Worker loading errors: Now logged with specific reasons
   - importScripts wrapped in try-catch
   - Posts error message to main thread
   - Verifies transformers object exists after load

Testing:
- TTS Preview should now work when clicking Settings tab
- Browser Whisper should load from CDN (or show specific error)
- Console logs will show exact init sequence
2026-03-31 16:00:10 +00:00
ifedan-ed
d96a008dfe v15: Fix TTS preview + Browser Whisper preload with extensive debugging
BREAKING FIXES:
- TTS Preview: Added event.preventDefault(), console logging, proper init check
- Browser Whisper: Complete console logging pipeline, error handling, progress tracking
- Voice Preferences: DOMContentLoaded fallback, explicit button click handlers
- Whisper Worker: Console logs at every step, better error messages

Debugging Features:
- Console logs show: button clicks, init events, progress updates, errors
- Progress tracking: [WhisperWorker] Progress: model.bin 47%
- Error messages: Specific failure reasons (not generic failures)
- Timeout warnings: 30s check for stuck downloads

Audio Backup Confirmed:
- Deletes immediately on successful transcription (line 621-624 app.js)
- NOT after 24 hours - 24h is server retention limit for failed transcriptions
- User was correct - this is working as designed

How to Debug:
1. Open DevTools → Console (F12)
2. Click button
3. Watch for [VoicePrefs] or [BrowserWhisper] logs
4. Check Network tab for actual downloads
5. Report what you see in console
2026-03-31 15:28:38 +00:00
Daniel Onyejesi
f126cf9fd7 Add browser-side Whisper transcription (local, zero network, HIPAA-safe)
- whisperWorker.js: Web Worker running @xenova/transformers Whisper in WASM
- browserWhisper.js: main-thread manager — audio→Float32 conversion, worker lifecycle
- transcribeAudio() checks BrowserWhisper.isEnabled() first, falls back to server
- Settings UI: enable/disable, model picker (tiny/base/small), pre-download button
- CSP: add wasm-unsafe-eval, cdn.jsdelivr.net, HuggingFace CDN domains
- Default: whisper-tiny.en (~39MB, ~2-3s per clip)
2026-03-31 07:30:17 -04:00