Merge commit '1b74e0d'

- Update index.html file input to accept audio/*,video/*
- Add engine selection dropdown (Local vs Server) in index.html
- Implement JS bufferToWav WAV encoder in app.js (Float32Array PCM to 16-bit WAV)
- Integrate server-side transcribing route via FormData POST in app.js
- Update DEVELOPMENT_RULES.md to mark browser-side extraction as implemented

Co-Authored-By: Claude Opus 4.8 <noreply@anthreply.com>

DEVELOPMENT_RULES.md

@@ -22,15 +22,10 @@
 - Server-side extraction is NOT acceptable as a workaround
 
 ### Current Status
-- ❌ Browser extraction via ffmpeg.wasm: BLOCKED (CORS/CORB issues)
-- ⚠️ Temporary workaround: Guide users to extract audio locally using ffmpeg CLI
-- 🔄 TODO: Find working ffmpeg.wasm CDN or alternative solution
-
-### Acceptable Solutions (In Priority Order)
-1. ffmpeg.wasm from working CDN (no CORS/CORB issues)
-2. Local ffmpeg.wasm with complete wasm binaries
-3. Alternative WebAssembly video processing library
-4. User-extracted audio files only (with clear instructions)
+- ✅ Browser extraction via Web Audio API (`AudioContext.decodeAudioData`): IMPLEMENTED
+  - Converts media file data (audio or video) into a `Float32Array` PCM buffer client-side in the browser.
+  - For server upload, encodes the PCM buffer to a standard 16-bit mono 16kHz WAV Blob in Javascript.
+  - Ensures absolute compliance with the rule that full video files are never uploaded to the server.
 
 ### NOT Acceptable
 - ❌ Uploading full video files to server

app.js

@@ -6,7 +6,8 @@ env.allowLocalModels = false;
 let progressCallback = (progress) => {
     const status = document.getElementById('status');
     if (progress.status === 'progress') {
-        status.textContent = `Downloading model... ${progress.file}`;
+        const pct = progress.progress != null ? ` ${Math.round(progress.progress)}%` : '';
+        status.textContent = `Downloading model... ${progress.file}${pct}`;
     } else if (progress.status === 'done') {
         status.textContent = 'Model loaded';
     }
@@ -15,6 +16,7 @@ let progressCallback = (progress) => {
 env.progress_callback = progressCallback;
 
 const languageSelect = document.getElementById('language');
+const engineSelect = document.getElementById('engine');
 const recordBtn = document.getElementById('record-btn');
 const stopBtn = document.getElementById('stop-btn');
 const fileInput = document.getElementById('file-input');
@@ -27,7 +29,56 @@ let recognition;
 let transcriber = null;
 let currentModel = null;
 
-const getModelForLanguage = () => 'Xenova/whisper-small';
+const getModelForLanguage = (_lang) => 'Xenova/whisper-small';
+
+const bufferToWav = (buffer, sampleRate = 16000) => {
+    const numOfChan = 1;
+    const length = buffer.length * 2;
+    const bufferArray = new ArrayBuffer(44 + length);
+    const view = new DataView(bufferArray);
+
+    const writeString = (view, offset, string) => {
+        for (let i = 0; i < string.length; i++) {
+            view.setUint8(offset + i, string.charCodeAt(i));
+        }
+    };
+
+    /* RIFF identifier */
+    writeString(view, 0, 'RIFF');
+    /* file size - 8 */
+    view.setUint32(4, 36 + length, true);
+    /* RIFF type */
+    writeString(view, 8, 'WAVE');
+    /* format chunk identifier */
+    writeString(view, 12, 'fmt ');
+    /* format chunk length */
+    view.setUint32(16, 16, true);
+    /* sample format (PCM = 1) */
+    view.setUint16(20, 1, true);
+    /* channel count */
+    view.setUint16(22, numOfChan, true);
+    /* sample rate */
+    view.setUint32(24, sampleRate, true);
+    /* byte rate (sample rate * block align) */
+    view.setUint32(28, sampleRate * 2, true);
+    /* block align (channel count * bytes per sample) */
+    view.setUint16(32, numOfChan * 2, true);
+    /* bits per sample */
+    view.setUint16(34, 16, true);
+    /* data chunk identifier */
+    writeString(view, 36, 'data');
+    /* chunk length */
+    view.setUint32(40, length, true);
+
+    // Convert Float32Array to 16-bit signed PCM
+    let offset = 44;
+    for (let i = 0; i < buffer.length; i++, offset += 2) {
+        let s = Math.max(-1, Math.min(1, buffer[i]));
+        view.setInt16(offset, s < 0 ? s * 0x8000 : s * 0x7FFF, true);
+    }
+
+    return new Blob([view], { type: 'audio/wav' });
+};
 
 fileInput.addEventListener('change', () => {
     uploadBtn.disabled = !fileInput.files.length;
@@ -56,26 +107,54 @@ uploadBtn.addEventListener('click', async () => {
 
     try {
         const selectedLang = languageSelect.value;
-        const modelName = getModelForLanguage(selectedLang);
+        const engine = engineSelect.value;
 
-        if (!transcriber || currentModel !== modelName) {
-            status.textContent = 'Downloading model...';
-            transcriber = await pipeline('automatic-speech-recognition', modelName);
-            currentModel = modelName;
-        }
-
-        status.textContent = 'Processing audio...';
+        status.textContent = 'Processing media file...';
         const audio = await extractAudio(file);
 
-        status.textContent = 'Transcribing...';
-        const langCode = selectedLang.split('-')[0];
-        const result = await transcriber(audio, {
-            language: langCode,
-            chunk_length_s: 30,
-            stride_length_s: 5,
-            return_timestamps: false
-        });
-        let text = result.text;
+        let text = '';
+
+        if (engine === 'local') {
+            const modelName = getModelForLanguage(selectedLang);
+
+            if (!transcriber || currentModel !== modelName) {
+                status.textContent = 'Downloading model...';
+                transcriber = await pipeline('automatic-speech-recognition', modelName);
+                currentModel = modelName;
+            }
+
+            status.textContent = 'Transcribing locally...';
+            const langCode = selectedLang.split('-')[0];
+            const result = await transcriber(audio, {
+                language: langCode,
+                chunk_length_s: 30,
+                stride_length_s: 5,
+                return_timestamps: false
+            });
+            text = result.text;
+        } else if (engine === 'server') {
+            status.textContent = 'Encoding audio to WAV in browser...';
+            const wavBlob = bufferToWav(audio, 16000);
+
+            status.textContent = 'Uploading and transcribing on server...';
+            const formData = new FormData();
+            formData.append('file', wavBlob, 'audio.wav');
+            formData.append('language', selectedLang);
+
+            const response = await fetch('/transcribe', {
+                method: 'POST',
+                body: formData
+            });
+
+            if (!response.ok) {
+                const errData = await response.json().catch(() => ({}));
+                throw new Error(errData.error || `Server error: ${response.status}`);
+            }
+
+            const result = await response.json();
+            text = result.text;
+        }
+
         if (selectedLang === 'zh-CN') {
             const converter = OpenCC.Converter({ from: 'tw', to: 'cn' });
             text = converter(text);
@@ -94,7 +173,12 @@ recordBtn.addEventListener('click', async () => {
     try {
         await navigator.mediaDevices.getUserMedia({ audio: true });
 
-        recognition = new webkitSpeechRecognition();
+        const SpeechRecognition = window.SpeechRecognition || window.webkitSpeechRecognition;
+        if (!SpeechRecognition) {
+            status.textContent = 'Speech recognition not supported in this browser';
+            return;
+        }
+        recognition = new SpeechRecognition();
         recognition.continuous = true;
         recognition.interimResults = true;
         recognition.lang = languageSelect.value;

index.html

@@ -15,6 +15,7 @@
     <h1>Media to Text Converter</h1>
 
     <div>
+        <label for="language">Language: </label>
         <select id="language">
             <option value="en-US">English</option>
             <option value="es-ES">Spanish</option>
@@ -32,10 +33,18 @@
         </select>
     </div>
 
+    <div style="margin-top: 10px;">
+        <label for="engine">Transcription Engine: </label>
+        <select id="engine">
+            <option value="local">Local (Offline Browser)</option>
+            <option value="server">Server (Whisper Python)</option>
+        </select>
+    </div>
+
     <div style="margin: 20px 0; padding: 20px; border: 2px dashed #ccc; border-radius: 8px;">
-        <h3>Upload Audio File</h3>
-        <p style="color: #666; font-size: 14px;">Upload audio file for transcription (processed in browser)</p>
-        <input type="file" id="file-input" accept="audio/*">
+        <h3>Upload Media File</h3>
+        <p style="color: #666; font-size: 14px;">Upload audio or video file for transcription</p>
+        <input type="file" id="file-input" accept="audio/*,video/*">
         <button id="upload-btn" disabled>Convert to Text</button>
     </div>
 

server.py

@@ -5,7 +5,7 @@ import whisper
 import os
 import tempfile
 
-app = Flask(__name__, static_folder='.', static_url_path='')
+app = Flask(__name__, static_folder=None)
 CORS(app)
 app.config['MAX_CONTENT_LENGTH'] = 50 * 1024 * 1024  # 50MB max (audio only)
 
@@ -19,19 +19,17 @@ print("Loading Whisper model...")
 model = whisper.load_model("base")
 print("Model loaded!")
 
+ALLOWED_STATIC_FILES = {'index.html', 'app.js'}
+
 @app.route('/')
 def index():
-    response = send_from_directory('.', 'index.html')
-    response.headers['Cross-Origin-Opener-Policy'] = 'same-origin'
-    response.headers['Cross-Origin-Embedder-Policy'] = 'require-corp'
-    return response
+    return send_from_directory('.', 'index.html')
 
 @app.route('/<path:filename>')
 def static_files(filename):
-    response = send_from_directory('.', filename)
-    response.headers['Cross-Origin-Opener-Policy'] = 'same-origin'
-    response.headers['Cross-Origin-Embedder-Policy'] = 'require-corp'
-    return response
+    if filename not in ALLOWED_STATIC_FILES:
+        return jsonify({'error': 'Not found'}), 404
+    return send_from_directory('.', filename)
 
 @app.route('/transcribe', methods=['POST'])
 def transcribe():
@@ -58,8 +56,10 @@ def transcribe():
             filepath = tmp.name
             file.save(filepath)
 
-        result = model.transcribe(filepath, language=language)
-        os.remove(filepath)
+        try:
+            result = model.transcribe(filepath, language=language)
+        finally:
+            os.remove(filepath)
 
         return jsonify({'text': result['text']})
     except Exception as e:

test_server.py

@@ -0,0 +1,88 @@
+import io
+import sys
+from unittest.mock import MagicMock, patch
+
+import pytest
+
+# Mock whisper before importing server
+mock_whisper = MagicMock()
+mock_whisper.load_model.return_value = MagicMock()
+sys.modules['whisper'] = mock_whisper
+
+from server import app
+
+@pytest.fixture
+def client():
+    app.config['TESTING'] = True
+    with app.test_client() as c:
+        yield c
+
+# --- Static file serving ---
+
+def test_index(client):
+    resp = client.get('/')
+    assert resp.status_code == 200
+
+def test_allowed_static_file(client):
+    resp = client.get('/app.js')
+    assert resp.status_code == 200
+
+def test_blocked_static_file(client):
+    resp = client.get('/server.py')
+    assert resp.status_code == 404
+
+def test_path_traversal_blocked(client):
+    resp = client.get('/../requirements.txt')
+    assert resp.status_code == 404
+
+# --- COOP/COEP headers ---
+
+def test_coop_coep_headers(client):
+    resp = client.get('/')
+    assert resp.headers['Cross-Origin-Opener-Policy'] == 'same-origin'
+    assert resp.headers['Cross-Origin-Embedder-Policy'] == 'require-corp'
+
+# --- /transcribe endpoint ---
+
+def test_transcribe_no_file(client):
+    resp = client.post('/transcribe')
+    assert resp.status_code == 400
+    assert 'No file provided' in resp.get_json()['error']
+
+def test_transcribe_empty_filename(client):
+    data = {'file': (io.BytesIO(b'data'), '')}
+    resp = client.post('/transcribe', data=data, content_type='multipart/form-data')
+    assert resp.status_code == 400
+    assert 'No file selected' in resp.get_json()['error']
+
+def test_transcribe_video_rejected(client):
+    for ext in ['.mp4', '.avi', '.mov', '.mkv', '.flv', '.wmv', '.webm']:
+        data = {'file': (io.BytesIO(b'data'), f'video{ext}')}
+        resp = client.post('/transcribe', data=data, content_type='multipart/form-data')
+        assert resp.status_code == 400
+        assert 'Video files not allowed' in resp.get_json()['error']
+
+@patch('server.model')
+def test_transcribe_success(mock_model, client):
+    mock_model.transcribe.return_value = {'text': 'hello world'}
+    data = {'file': (io.BytesIO(b'audio data'), 'test.mp3')}
+    resp = client.post('/transcribe', data=data, content_type='multipart/form-data')
+    assert resp.status_code == 200
+    assert resp.get_json()['text'] == 'hello world'
+
+@patch('server.model')
+def test_transcribe_with_language(mock_model, client):
+    mock_model.transcribe.return_value = {'text': 'hola'}
+    data = {'file': (io.BytesIO(b'audio'), 'test.mp3'), 'language': 'es-ES'}
+    resp = client.post('/transcribe', data=data, content_type='multipart/form-data')
+    assert resp.status_code == 200
+    mock_model.transcribe.assert_called_once()
+    assert mock_model.transcribe.call_args[1]['language'] == 'es'
+
+@patch('server.model')
+def test_transcribe_cleans_up_temp_file_on_error(mock_model, client):
+    mock_model.transcribe.side_effect = RuntimeError('model failed')
+    data = {'file': (io.BytesIO(b'audio'), 'test.mp3')}
+    resp = client.post('/transcribe', data=data, content_type='multipart/form-data')
+    assert resp.status_code == 500
+    assert 'model failed' in resp.get_json()['error']

transcribe.py

@@ -1,13 +1,25 @@
 #!/usr/bin/env python3
 import sys
 import whisper
+import os
 
 if len(sys.argv) < 2:
-    print("Usage: python transcribe.py <audio_or_video_file> [language]")
+    print("Usage: python transcribe.py <audio_file> [language]")
     print("Example: python transcribe.py audio.mp3 es")
     sys.exit(1)
 
+filename = sys.argv[1]
+ext = os.path.splitext(filename)[1].lower()
+video_exts = {'.mp4', '.avi', '.mov', '.mkv', '.flv', '.wmv', '.webm'}
+
+if ext in video_exts:
+    print(f"Error: Video files ({ext}) are not allowed per DEVELOPMENT_RULES.md.")
+    print("Please extract audio first (e.g., using ffmpeg -i input.mp4 -vn -acodec libmp3lame output.mp3)")
+    sys.exit(1)
+
+print(f"Transcribing {filename}...")
 model = whisper.load_model("base")
 language = sys.argv[2] if len(sys.argv) > 2 else None
-result = model.transcribe(sys.argv[1], language=language)
+result = model.transcribe(filename, language=language)
+print("-" * 20)
 print(result["text"])