Update app.py

app.py

@@ -1,29 +1,54 @@
+import os
+import subprocess
+import sys
+
+# Forcer l'installation de llama-cpp-python pré-compilé avant de lancer le reste
+try:
+    import llama_cpp
+except ImportError:
+    print("Installation de llama-cpp-python...")
+    subprocess.check_call([
+        sys.executable, "-m", "pip", "install", 
+        "llama-cpp-python", 
+        "--extra-index-url", "https://abetlen.github.io/llama-cpp-python/whl/cpu"
+    ])
+    import llama_cpp
+
 import gradio as gr
-from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 
-# Téléchargement de ton modèle
-model_path = hf_hub_download(
-    repo_id="AboyNight/KryZen", 
-    filename="Dolphin3.0-Llama3.1-8B-Q5_K_M.gguf"
-)
+# Configuration du modèle
+REPO_ID = "AboyNight/KryZen"
+FILENAME = "Dolphin3.0-Llama3.1-8B-Q5_K_M.gguf"
 
-# Chargement du modèle
-llm = Llama(model_path=model_path, n_ctx=2048, n_threads=4)
+print(f"Téléchargement du modèle {FILENAME}...")
+model_path = hf_hub_download(repo_id=REPO_ID, filename=FILENAME)
+
+print("Chargement du modèle en mémoire...")
+llm = llama_cpp.Llama(
+    model_path=model_path,
+    n_ctx=2048,
+    n_threads=4  # Adapté au CPU gratuit de HF
+)
 
 def generate(message, history):
-    # Format spécial pour Dolphin 3.0 (ChatML)
+    # Format ChatML pour Dolphin 3.0
     prompt = f"<|im_start|>user\n{message}<|im_end|>\n<|im_start|>assistant\n"
     
     output = llm(
-        prompt, 
-        max_tokens=512, 
-        stop=["<|im_end|>", "<|im_start|>"], 
+        prompt,
+        max_tokens=512,
+        stop=["<|im_end|>", "<|im_start|>"],
         echo=False
     )
     return output["choices"][0]["text"]
 
-demo = gr.ChatInterface(fn=generate, title="KryZen AI (Dolphin 3.0)")
+# Interface Gradio
+demo = gr.ChatInterface(
+    fn=generate, 
+    title="KryZen AI (Dolphin 3.0)",
+    description="Hébergé gratuitement sur HuggingFace Spaces"
+)
 
 if __name__ == "__main__":
     demo.launch(server_name="0.0.0.0", server_port=7860)