1. Only download the specific `.gguf` file we need

2. Use less memory during download
3. Avoid the "Killed" error from system memory exhaustion

src/app.py

@@ -81,30 +81,26 @@ MODEL_DIR = os.path.join(BASE_DIR, "models")
 MODEL_PATH = os.path.join(MODEL_DIR, MODEL_NAME)
 
 def ensure_model():
+    """Download specific model file if not exists."""
     # Create models/ directory if missing
     os.makedirs(MODEL_DIR, exist_ok=True)
     
-    # Download model if it's not already there
     model_path = os.path.join(MODEL_DIR, MODEL_NAME)
     if not os.path.isfile(model_path):
-        print(f"Downloading model from {REPO_ID}...")
-        # Download to a subdirectory to avoid file conflicts
-        download_dir = os.path.join(MODEL_DIR, "download_cache")
-        snapshot_download(
-            repo_id=REPO_ID,
-            repo_type="model",
-            local_dir=download_dir,
-            local_dir_use_symlinks=False
-        )
+        print(f"Downloading model {MODEL_NAME} from {REPO_ID}...")
         
-        # Move the specific model file we want to models/
-        src_path = os.path.join(download_dir, MODEL_NAME)
-        if os.path.exists(src_path):
-            import shutil
-            shutil.move(src_path, model_path)
-            print(f"Moved model to {model_path}")
-        else:
-            raise ValueError(f"Downloaded files but couldn't find {MODEL_NAME}")
+        # Download specific file instead of whole repo
+        from huggingface_hub import hf_hub_download
+        try:
+            hf_hub_download(
+                repo_id=REPO_ID,
+                filename=MODEL_NAME,
+                local_dir=MODEL_DIR,
+                local_dir_use_symlinks=False
+            )
+            print(f"Successfully downloaded model to {model_path}")
+        except Exception as e:
+            raise RuntimeError(f"Failed to download model: {str(e)}")
     else:
         print(f"Model already exists at {model_path}")