fix

backend_service.py

@@ -2,15 +2,6 @@
 FastAPI Backend AI Service converted from Gradio app
 Provides OpenAI-compatible chat completion endpoints
 """
-# Configure Hugging Face cache directory and authentication before any HF imports
-import os
-_CACHE_DIR = os.path.join(os.getcwd(), ".hf_cache")
-os.makedirs(_CACHE_DIR, exist_ok=True)
-# Set environment variables early to avoid default ~/.cache permission issues
-os.environ.setdefault("HF_HOME", _CACHE_DIR)
-os.environ.setdefault("TRANSFORMERS_CACHE", _CACHE_DIR)
-# Authentication token for private models
-hf_token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_TOKEN")
 
 import asyncio
 import logging
@@ -41,9 +32,6 @@ except ImportError:
 logging.basicConfig(level=logging.INFO)
 logger = logging.getLogger(__name__)
 
-# Authentication token for private models
-hf_token = os.getenv("HF_TOKEN") or os.getenv("HUGGINGFACE_TOKEN")
-
 # Pydantic models for multimodal content
 class TextContent(BaseModel):
     type: str = Field(default="text", description="Content type")
@@ -176,35 +164,21 @@ async def lifespan(app: FastAPI):
     """Application lifespan manager for startup and shutdown events"""
     global inference_client, tokenizer, image_text_pipeline
     
-    logger.info(f"Using Hugging Face cache directory: {_CACHE_DIR}")
-    if not hf_token:
-        logger.info("No Hugging Face auth token provided; private models may fail to load.")
-
     # Startup
     logger.info("🚀 Starting AI Backend Service...")
     try:
         # Initialize HuggingFace Inference Client for text generation
-        inference_client = InferenceClient(model=current_model, token=hf_token)
+        inference_client = InferenceClient(model=current_model)
         logger.info(f"✅ Initialized inference client with model: {current_model}")
         
         # Initialize image-text-to-text pipeline
         if transformers_available and pipeline:
             try:
                 logger.info(f"🖼️ Initializing image captioning pipeline with model: {vision_model}")
-                image_text_pipeline = pipeline(
-                    "image-to-text",
-                    model=vision_model,
-                    use_auth_token=hf_token,
-                    cache_dir=_CACHE_DIR
-                )
+                image_text_pipeline = pipeline("image-to-text", model=vision_model)  # Use image-to-text task
                 logger.info("✅ Image captioning pipeline loaded successfully")
-            except PermissionError as pe:
-                logger.warning(f"⚠️ Permission error while loading image captioning pipeline: {pe}. Check cache directory permissions: {_CACHE_DIR}")
-                image_text_pipeline = None
             except Exception as e:
                 logger.warning(f"⚠️ Could not load image captioning pipeline: {e}")
-                if "not a local folder and is not a valid model identifier" in str(e):
-                    logger.warning("Model identifier invalid; ensure model name is correct or you have access.")
                 image_text_pipeline = None
         else:
             logger.warning("⚠️ Transformers not available, image processing disabled")
@@ -213,24 +187,13 @@ async def lifespan(app: FastAPI):
         # Initialize tokenizer for better text handling
         if transformers_available and AutoTokenizer:
             try:
-                tokenizer = AutoTokenizer.from_pretrained(
-                    current_model,
-                    use_auth_token=hf_token,
-                    cache_dir=_CACHE_DIR
-                )  # type: ignore
+                tokenizer = AutoTokenizer.from_pretrained(current_model)  # type: ignore
                 logger.info("✅ Tokenizer loaded successfully")
-            except PermissionError as pe:
-                logger.warning(f"⚠️ Permission error while loading tokenizer: {pe}. Check cache directory permissions: {_CACHE_DIR}")
-                tokenizer = None
-            except ValueError as ve:
-                logger.warning(f"⚠️ Could not load tokenizer: {ve}. If this is a private model, set HF_TOKEN env var to access it.")
-                tokenizer = None
             except Exception as e:
                 logger.warning(f"⚠️ Could not load tokenizer: {e}")
                 tokenizer = None
         else:
             logger.info("⚠️ Tokenizer initialization skipped")
-            tokenizer = None
             
     except Exception as e:
         logger.error(f"❌ Failed to initialize inference client: {e}")