deploy

Dockerfile

@@ -24,8 +24,8 @@ COPY app.py .
 # Set environment variable for Hugging Face cache
 ENV HF_HOME=/app/cache
 
-# Expose port
+# Expose default port (can be overridden by PORT env var)
 EXPOSE 8000
 
-# Run the FastAPI app with uvicorn
-CMD ["sh", "-c", "uvicorn app:app --host 0.0.0.0 --port ${PORT:-8000}"]
+# Run the FastAPI app with uvicorn, using PORT env var
+CMD ["sh", "-c", "uvicorn app:app --host 0.0.0.0 --port ${PORT:-8000} --workers 1"]

app.py

@@ -10,6 +10,7 @@ import subprocess
 import tempfile
 from contextlib import contextmanager
 
+# Configure logging
 logging.basicConfig(
     level=logging.INFO,
     format="%(asctime)s [%(levelname)s] %(message)s",
@@ -20,6 +21,7 @@ logging.basicConfig(
 )
 logger = logging.getLogger(__name__)
 
+# Log system information
 logger.info(f"Python version: {sys.version}")
 try:
     import transformers
@@ -30,26 +32,33 @@ except ImportError as e:
     logger.error(f"Failed to import dependency: {str(e)}")
     raise
 
+# Set up cache directory
 os.makedirs("/app/cache", exist_ok=True)
 os.environ["HF_HOME"] = "/app/cache"
 logger.info(f"Set HF_HOME to /app/cache")
 
+# Initialize FastAPI app
 app = FastAPI(title="Quran Transcription API")
 
+# Health check endpoint
 @app.get("/health", status_code=200)
 async def health_check():
     logger.info("Health check requested")
     return {"status": "healthy", "model_loaded": model is not None}
 
+# Debug endpoint
 @app.get("/debug")
 async def debug():
+    logger.info("Debug endpoint requested")
     return {
         "cuda_available": torch.cuda.is_available(),
         "model_loaded": model is not None,
         "pipeline_initialized": asr is not None,
-        "cache_dir": os.getenv("HF_HOME")
+        "cache_dir": os.getenv("HF_HOME"),
+        "port": os.getenv("PORT", "8000")
     }
 
+# Load model and processor
 try:
     model_id = "tarteel-ai/whisper-base-ar-quran"
     logger.info(f"Loading processor for model: {model_id}")
@@ -61,6 +70,7 @@ except Exception as e:
     logger.error(f"Failed to load model: {str(e)}")
     raise HTTPException(status_code=500, detail="Model loading failed")
 
+# Initialize ASR pipeline
 try:
     logger.info("Initializing ASR pipeline")
     asr = pipeline(
@@ -68,7 +78,7 @@ try:
         model=model,
         tokenizer=processor.tokenizer,
         feature_extractor=processor.feature_extractor,
-        device=-1  # Force CPU for testing
+        device=-1  # Force CPU
     )
 except Exception as e:
     logger.error(f"Failed to initialize ASR pipeline: {str(e)}")
@@ -111,7 +121,8 @@ async def transcribe_audio(file: UploadFile = File(...)):
                 ["ffmpeg", "-i", temp_mp3.name, "-ar", "16000", "-ac", "1", "-y", temp_wav_path],
                 check=True,
                 capture_output=True,
-                text=True
+                text=True,
+                timeout=30
             )
             logger.debug(f"ffmpeg output: {result.stdout}")
         except subprocess.CalledProcessError as e:
@@ -145,11 +156,12 @@ async def startup_event():
     logger.info(f"Memory allocated: {torch.cuda.memory_allocated() if torch.cuda.is_available() else 'N/A'}")
 
 if __name__ == "__main__":
-    port = int(os.getenv("PORT", 8000))
+    port = int(os.getenv("PORT", 8000))  # Use PORT env var or default to 8000
     logger.info(f"Starting Uvicorn server on port {port}")
     uvicorn.run(
         "app:app",
         host="0.0.0.0",
         port=port,
-        log_level="info"
+        log_level="info",
+        workers=1  # Single worker to avoid resource issues
     )