Auto-deploy from GitHub: 07597185080a5be8bd74c45bc69a05d84f5e6604

Dockerfile

@@ -25,10 +25,5 @@ RUN mkdir -p uploads temp_dir
 # Expose port
 EXPOSE 7860
 
-# Create a startup script
-RUN echo '#!/bin/bash\n\
-python worker.py &\n\
-python app.py' > /app/start.sh && chmod +x /app/start.sh
-
-# Run the application
-CMD ["/app/start.sh"]
+# Run only the Flask app (worker starts automatically on first upload)
+CMD ["python", "app.py"]

app.py

@@ -5,6 +5,9 @@ import os
 import uuid
 from datetime import datetime
 from werkzeug.utils import secure_filename
+import threading
+import subprocess
+import time
 
 app = Flask(__name__)
 CORS(app)
@@ -15,6 +18,10 @@ ALLOWED_EXTENSIONS = {'wav', 'mp3', 'flac', 'ogg', 'm4a', 'aac'}
 os.makedirs(UPLOAD_FOLDER, exist_ok=True)
 os.makedirs('temp_dir', exist_ok=True)
 
+# Worker state
+worker_thread = None
+worker_running = False
+
 def init_db():
     conn = sqlite3.connect('audio_captions.db')
     c = conn.cursor()
@@ -32,6 +39,129 @@ def init_db():
 def allowed_file(filename):
     return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
 
+def start_worker():
+    """Start the worker thread if not already running"""
+    global worker_thread, worker_running
+    
+    if not worker_running:
+        worker_running = True
+        worker_thread = threading.Thread(target=worker_loop, daemon=True)
+        worker_thread.start()
+        print("✅ Worker thread started")
+
+def worker_loop():
+    """Main worker loop that processes audio files"""
+    print("🤖 STT Worker started. Monitoring for new audio files...")
+    
+    CWD = "./"
+    PYTHON_PATH = "stt-transcribe"
+    STT_MODEL_NAME = "fasterwhispher"
+    POLL_INTERVAL = 3  # seconds
+    
+    import shlex
+    import json
+    
+    while worker_running:
+        try:
+            # Get next unprocessed file
+            conn = sqlite3.connect('audio_captions.db')
+            conn.row_factory = sqlite3.Row
+            c = conn.cursor()
+            c.execute('''SELECT * FROM audio_files 
+                         WHERE status = 'not_started' 
+                         ORDER BY created_at ASC 
+                         LIMIT 1''')
+            row = c.fetchone()
+            conn.close()
+            
+            if row:
+                file_id = row['id']
+                filepath = row['filepath']
+                filename = row['filename']
+                
+                print(f"\n{'='*60}")
+                print(f"🎵 Processing: {filename}")
+                print(f"📝 ID: {file_id}")
+                print(f"{'='*60}")
+                
+                # Update status to processing
+                update_status(file_id, 'processing')
+                
+                try:
+                    # Run STT command
+                    print(f"🔄 Running STT on: {os.path.abspath(filepath)}")
+                    command = f"""cd {CWD} && {PYTHON_PATH} --input {shlex.quote(os.path.abspath(filepath))} --model {STT_MODEL_NAME}"""
+                    
+                    subprocess.run(
+                        command,
+                        shell=True,
+                        executable="/bin/bash",
+                        check=True,
+                        cwd=CWD,
+                        env={
+                            **os.environ,
+                            'PYTHONUNBUFFERED': '1',
+                            'CUDA_LAUNCH_BLOCKING': '1',
+                            'USE_CPU_IF_POSSIBLE': 'true'
+                        }
+                    )
+                    
+                    # Read transcription result
+                    output_path = f'{CWD}/temp_dir/output_transcription.json'
+                    with open(output_path, 'r') as file:
+                        result = json.loads(file.read().strip())
+                    
+                    # Extract caption text
+                    caption = result.get('text', '') or result.get('transcription', '') or str(result)
+                    
+                    print(f"✅ Successfully processed: {filename}")
+                    print(f"📄 Caption preview: {caption[:100]}...")
+                    
+                    # Update database with success
+                    update_status(file_id, 'completed', caption=caption)
+                    
+                    # Delete the audio file after successful processing
+                    if os.path.exists(filepath):
+                        os.remove(filepath)
+                        print(f"🗑️  Deleted audio file: {filepath}")
+                    
+                except Exception as e:
+                    print(f"❌ Failed to process: {filename}")
+                    print(f"Error: {str(e)}")
+                    update_status(file_id, 'failed', error=str(e))
+                    
+                    # Don't delete file on failure (for debugging)
+                    # Optionally delete after some time or manual review
+                    
+            else:
+                # No files to process, sleep for a bit
+                time.sleep(POLL_INTERVAL)
+                
+        except Exception as e:
+            print(f"⚠️  Worker error: {str(e)}")
+            time.sleep(POLL_INTERVAL)
+
+def update_status(file_id, status, caption=None, error=None):
+    """Update the status of a file in the database"""
+    conn = sqlite3.connect('audio_captions.db')
+    c = conn.cursor()
+    
+    if status == 'completed':
+        c.execute('''UPDATE audio_files 
+                     SET status = ?, caption = ?, processed_at = ?
+                     WHERE id = ?''',
+                  (status, caption, datetime.now().isoformat(), file_id))
+    elif status == 'failed':
+        c.execute('''UPDATE audio_files 
+                     SET status = ?, caption = ?, processed_at = ?
+                     WHERE id = ?''',
+                  (status, f"Error: {error}", datetime.now().isoformat(), file_id))
+    else:
+        c.execute('UPDATE audio_files SET status = ? WHERE id = ?', (status, file_id))
+    
+    conn.commit()
+    conn.close()
+
 @app.route('/')
 def index():
     return send_from_directory('.', 'index.html')
@@ -63,6 +193,9 @@ def upload_audio():
     conn.commit()
     conn.close()
     
+    # Start worker on first upload
+    start_worker()
+    
     return jsonify({
         'id': file_id,
         'filename': filename,
@@ -115,10 +248,21 @@ def get_file(file_id):
 
 @app.route('/health', methods=['GET'])
 def health():
-    return jsonify({'status': 'healthy', 'service': 'audio-caption-generator'})
+    return jsonify({
+        'status': 'healthy',
+        'service': 'audio-caption-generator',
+        'worker_running': worker_running
+    })
 
 if __name__ == '__main__':
     init_db()
+    print("\n" + "="*60)
+    print("🚀 Audio Caption Generator API Server")
+    print("="*60)
+    print("📌 Worker will start automatically on first upload")
+    print("🗑️  Audio files will be deleted after successful processing")
+    print("="*60 + "\n")
+    
     # Use PORT environment variable for Hugging Face compatibility
     port = int(os.environ.get('PORT', 7860))
     app.run(debug=False, host='0.0.0.0', port=port)

worker.py

@@ -9,12 +9,12 @@ from datetime import datetime
 CWD = "./"
 PYTHON_PATH = "stt-transcribe"
 STT_MODEL_NAME = "fasterwhispher"
-POLL_INTERVAL = 5  # seconds
+POLL_INTERVAL = 3  # seconds
 
 def process_audio(file_id, filepath):
     """Process audio file using STT and return the transcription"""
     try:
-        print(f"Processing file: {filepath}")
+        print(f"🔄 Running STT on: {os.path.abspath(filepath)}")
         
         # Run STT command
         command = f"""cd {CWD} && {PYTHON_PATH} --input {shlex.quote(os.path.abspath(filepath))} --model {STT_MODEL_NAME}"""
@@ -44,7 +44,7 @@ def process_audio(file_id, filepath):
         return caption, None
         
     except Exception as e:
-        print(f"Error processing file {file_id}: {str(e)}")
+        print(f"❌ Error processing file {file_id}: {str(e)}")
         return None, str(e)
 
 def update_status(file_id, status, caption=None, error=None):
@@ -70,7 +70,8 @@ def update_status(file_id, status, caption=None, error=None):
 
 def worker_loop():
     """Main worker loop that processes audio files"""
-    print("STT Worker started. Polling for new audio files...")
+    print("🤖 STT Worker started. Monitoring for new audio files...")
+    print("🗑️  Audio files will be deleted after successful processing\n")
     
     while True:
         try:
@@ -91,8 +92,8 @@ def worker_loop():
                 filename = row['filename']
                 
                 print(f"\n{'='*60}")
-                print(f"Processing: {filename}")
-                print(f"ID: {file_id}")
+                print(f"🎵 Processing: {filename}")
+                print(f"📝 ID: {file_id}")
                 print(f"{'='*60}")
                 
                 # Update status to processing
@@ -102,24 +103,36 @@ def worker_loop():
                 caption, error = process_audio(file_id, filepath)
                 
                 if caption:
-                    print(f"✓ Successfully processed: {filename}")
-                    print(f"Caption: {caption[:100]}...")
+                    print(f"✅ Successfully processed: {filename}")
+                    print(f"📄 Caption preview: {caption[:100]}...")
                     update_status(file_id, 'completed', caption=caption)
+                    
+                    # Delete the audio file after successful processing
+                    if os.path.exists(filepath):
+                        os.remove(filepath)
+                        print(f"🗑️  Deleted audio file: {filepath}")
                 else:
-                    print(f"✗ Failed to process: {filename}")
+                    print(f"❌ Failed to process: {filename}")
                     print(f"Error: {error}")
                     update_status(file_id, 'failed', error=error)
+                    # Don't delete file on failure (for debugging)
             else:
                 # No files to process, sleep for a bit
                 time.sleep(POLL_INTERVAL)
                 
         except Exception as e:
-            print(f"Worker error: {str(e)}")
+            print(f"⚠️  Worker error: {str(e)}")
             time.sleep(POLL_INTERVAL)
 
 if __name__ == '__main__':
     # Initialize database if it doesn't exist
     if not os.path.exists('audio_captions.db'):
-        print("Database not found. Please run app.py first to initialize.")
+        print("❌ Database not found. Please run app.py first to initialize.")
     else:
+        print("\n" + "="*60)
+        print("🚀 Starting STT Worker (Standalone Mode)")
+        print("="*60)
+        print("⚠️  Note: Worker is now embedded in app.py")
+        print("⚠️  This standalone mode is for testing/debugging only")
+        print("="*60 + "\n")
         worker_loop()