Update app.py

app.py

@@ -1,10 +1,8 @@
-
 import gradio as gr
 from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 
 try:
-    # This downloads the 9.6GB file into the Space's runtime temporary storage at startup
     print("--- Downloading Model from Model Hub ---")
     model_path = hf_hub_download(
         repo_id='Arabic250/gemma-4-gguf-export',
@@ -15,13 +13,23 @@ try:
     llm = Llama(model_path=model_path, n_ctx=2048, n_threads=2)
 
     def chat(message, history):
-        output = llm(f"USER: {message}\nASSISTANT: ", max_tokens=512, stop=['USER:'])
+        # بناء سياق المحادثة ليحتفظ بالرسائل السابقة
+        prompt = ""
+        for user_msg, bot_msg in history:
+            prompt += f"USER: {user_msg}\nASSISTANT: {bot_msg}\n"
+        
+        prompt += f"USER: {message}\nASSISTANT: "
+
+        # echo=False تمنع النموذج من تكرار السؤال في الإجابة
+        output = llm(prompt, max_tokens=512, stop=['USER:'], echo=False)
         return output['choices'][0]['text']
 
     demo = gr.ChatInterface(fn=chat, title='Gemma 4 Medical')
+
 except Exception as e:
     with gr.Blocks() as demo:
         gr.Markdown(f'# Error: {e}')
 
 if __name__ == '__main__':
-    demo.launch()
+    # تحديد المنفذ 7860 صراحةً وهو المنفذ الافتراضي الذي تستمع إليه Hugging Face Spaces
+    demo.launch(server_name="0.0.0.0", server_port=7860)