changed order for user permissions on checking model from cache tmp

Dockerfile

@@ -1,6 +1,12 @@
-FROM python:3.9-slim
+FROM python:3.11-slim
 
+# Create user first
 RUN useradd -m -u 1000 user
+
+# Create cache directory and set permissions while still root
+RUN mkdir -p /tmp/fasttext_cache && chown -R user:user /tmp/fasttext_cache
+
+# Switch to user after setting up permissions
 USER user
 ENV PATH="/home/user/.local/bin:$PATH"
 ENV MODEL_URL="https://huggingface.co/Miroir/cc.fr.300.reduced/resolve/main/cc.fr.300.reduced.vec"
@@ -9,7 +15,7 @@ WORKDIR /app
 
 COPY --chown=user requirements.txt requirements.txt
 
-# Modified installation command with --no-cache-dir
+# Install dependencies
 RUN pip install --no-cache-dir --upgrade pip && \
     pip install --no-cache-dir --upgrade -r requirements.txt
 

services/word_service.py

@@ -23,26 +23,27 @@ class WordEmbeddingService:
     def _initialize_model(self):
         """Initialize the model only when needed"""
         try:
-            # Get model URL from environment variable
             model_url = os.getenv('MODEL_URL', 'https://huggingface.co/Miroir/cc.fr.300.reduced/resolve/main/cc.fr.300.reduced.vec')
+            cache_dir = "/tmp/fasttext_cache"  # Hugging Face Spaces preserves this
+            os.makedirs(cache_dir, exist_ok=True)
             
-            logger.info("Loading FastText embeddings from URL...")
+            cache_path = os.path.join(cache_dir, "model.vec")
             
-            # Create a temporary file to store the model
-            with tempfile.NamedTemporaryFile(delete=False) as temp_file:
-                # Download the file
+            # Check if cached
+            if not os.path.exists(cache_path):
+                logger.info("Downloading FastText embeddings...")
                 response = requests.get(model_url, stream=True)
                 response.raise_for_status()
                 
-                # Write the content to the temporary file
-                for chunk in response.iter_content(chunk_size=8192):
-                    if chunk:
-                        temp_file.write(chunk)
-                
-                temp_file.flush()
-                
-                # Load the model from the temporary file
-                WordEmbeddingService._model = KeyedVectors.load_word2vec_format(temp_file.name)
+                with open(cache_path, 'wb') as f:
+                    for chunk in response.iter_content(chunk_size=8192):
+                        if chunk:
+                            f.write(chunk)
+            else:
+                logger.info("Using cached FastText model")
+            
+            # Load the model
+            WordEmbeddingService._model = KeyedVectors.load_word2vec_format(cache_path)
             
             # Build vocabulary vectors
             self.vocab_vectors = {
@@ -50,8 +51,7 @@ class WordEmbeddingService:
                 for word in WordEmbeddingService._model.index_to_key
             }
             
-            logger.info(f"FastText model loaded successfully with "
-                       f"{len(self.vocab_vectors)} words in the vocabulary.")
+            logger.info(f"FastText model loaded with {len(self.vocab_vectors)} words")
                        
         except Exception as e:
             logger.exception(f"Failed to load FastText model: {str(e)}")