Update app.py

app.py

@@ -7,20 +7,31 @@ from typing import Optional
 import re
 from functools import lru_cache
 import asyncio
-import logging.handlers
 
 # Set environment variables for Hugging Face cache
 os.environ["HF_HOME"] = "/app/cache"
 os.environ["TRANSFORMERS_CACHE"] = "/app/cache"
 
+# Environment configuration
+USE_8BIT = False
+try:
+    import bitsandbytes  # hanya untuk memastikan modul tersedia
+    USE_8BIT = os.getenv("USE_QUANTIZATION", "0") == "1"
+except ImportError:
+    USE_8BIT = False
+
+DEVICE = int(os.getenv("DEVICE", "-1"))  # -1 for CPU, 0+ for GPU
+MAX_TEXT_LENGTH = int(os.getenv("MAX_TEXT_LENGTH", "5000"))
+
 app = FastAPI()
 
-# Configure asynchronous logging with RotatingFileHandler and StreamHandler
+# Configure logging with timestamp and level
 logging.basicConfig(
+    format="%(asctime)s - %(name)s - %(levelname)s - %(message)s",
     level=logging.INFO,
     handlers=[
         logging.handlers.RotatingFileHandler("/app/logs/app.log", maxBytes=1000000, backupCount=1),
-        logging.StreamHandler()  # Added for console output
+        logging.StreamHandler()
     ]
 )
 logger = logging.getLogger(__name__)
@@ -43,13 +54,17 @@ def get_translator(lang: str):
     """Load or retrieve cached translator for the given language."""
     if lang not in translators:
         logger.info(f"Loading model for {lang} from {MODEL_MAP[lang]}...")
-        translators[lang] = pipeline(
-            "translation",
-            model=MODEL_MAP[lang],
-            device=-1,  # Explicitly use CPU for Hugging Face Spaces (free tier)
-            model_kwargs={"load_in_8bit": True} if os.getenv("USE_QUANTIZATION", "0") == "1" else {}
-        )
-        logger.info(f"Model for {lang} loaded successfully.")
+        try:
+            translators[lang] = pipeline(
+                "translation",
+                model=MODEL_MAP[lang],
+                device=DEVICE,
+                model_kwargs={"load_in_8bit": USE_8BIT}
+            )
+            logger.info(f"Model for {lang} loaded successfully.")
+        except Exception as e:
+            logger.error(f"Failed to load model for {lang}: {str(e)}")
+            raise
     return translators[lang]
 
 @lru_cache(maxsize=100)
@@ -96,6 +111,12 @@ async def translate(text: str, source_lang_override: Optional[str] = None):
     if not text:
         raise HTTPException(status_code=400, detail="Text input is required.")
 
+    if len(text) > MAX_TEXT_LENGTH:
+        raise HTTPException(
+            status_code=413,
+            detail=f"Text too long. Max allowed length: {MAX_TEXT_LENGTH}."
+        )
+
     try:
         # Determine source language
         if source_lang_override and source_lang_override in MODEL_MAP: