Update app.py

app.py

@@ -64,25 +64,20 @@ def get_llm():
     global _llm
     if _llm is not None:
         return _llm
-    if not _LLAMA_OK:
-        print("[law-app] llama_cpp non disponible.")
+    if not _LLAMA_AVAILABLE:
         return None
     with _llm_lock:
         if _llm is not None:
             return _llm
-        model_path = download_model()
-        if not model_path or not os.path.exists(model_path):
-            print("[law-app] Modèle introuvable.")
+        gguf_files = glob.glob("/app/*.gguf") + glob.glob("*.gguf")
+        if not gguf_files:
+            print("[spelling-app] No .gguf file found – AI tips disabled.")
             return None
-        print(f"[law-app] Chargement : {model_path}")
-        _llm = Llama(
-            model_path=model_path,
-            n_ctx=1024,
-            n_threads=int(os.getenv("N_THREADS", "2")),
-            n_gpu_layers=int(os.getenv("N_GPU_LAYERS", "0")),
-            verbose=False,
-        )
-        print("[law-app] Modèle prêt.")
+        model_path = gguf_files[0]
+        print(f"[spelling-app] Loading model: {model_path}")
+        _llm = Llama(model_path=model_path, n_ctx=512, n_threads=2,
+                     n_gpu_layers=int(os.getenv("N_GPU_LAYERS", "-1")), verbose=False)
+        print("[spelling-app] Model ready.")
         return _llm
 
 # ── Leçons ────────────────────────────────────────────────────────────────────