Update server.py

server.py

@@ -179,21 +179,31 @@ def chat(req: ChatRequest):
 
 os.makedirs("models", exist_ok=True)
 
-MODEL_URLS = {
-    "tinyllama": "https://huggingface.co/TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF/resolve/main/tinyllama-1.1b-chat.Q4_K_M.gguf",
-    "qwen": "https://huggingface.co/Qwen/Qwen1.5-1.8B-Chat-GGUF/resolve/main/qwen1_5-1_8b-chat-q4_k_m.gguf"
+MODEL_FILES = {
+    "tinyllama": (
+        "TheBloke/TinyLlama-1.1B-Chat-v1.0-GGUF",
+        "tinyllama-1.1b-chat.Q4_K_M.gguf"
+    ),
+    "qwen": (
+        "Qwen/Qwen1.5-1.8B-Chat-GGUF",
+        "qwen1_5-1_8b-chat-q4_k_m.gguf"
+    )
 }
 
-for name, url in MODEL_URLS.items():
-
+for name, (repo, file) in MODEL_FILES.items():
     path = f"models/{name}.gguf"
 
-    if not os.path.exists(path) or os.path.getsize(path) < 100000000:
-        print(f"Downloading model: {name}")
+    if not os.path.exists(path):
+        print(f"Downloading model {name} from {repo}")
+
+        downloaded = hf_hub_download(
+            repo_id=repo,
+            filename=file
+        )
 
-        os.system(f"curl -L {url} -o {path}")
+        os.system(f"cp {downloaded} {path}")
 
-        print(f"Downloaded {name}")
+        print(f"Model ready: {path}")
 
 # ---------------------------
 # Start server