Update app.py

app.py

@@ -6,14 +6,12 @@ from llama_cpp_agent.providers import LlamaCppPythonProvider
 from llama_cpp_agent.chat_history import BasicChatHistory
 from llama_cpp_agent.chat_history.messages import UserMessage, AssistantMessage
 
-# ⬇️ دانلود مدل
 hf_hub_download(
     repo_id="mradermacher/Arsh-llm-GGUF",
     filename="Arsh-llm.Q4_K_M.gguf",
     local_dir="./models"
 )
 
-# 🧠 لود مدل (فقط یکبار)
 llm = Llama(
     model_path="./models/Arsh-llm.Q4_K_M.gguf",
     n_batch=512,
@@ -25,14 +23,12 @@ provider = LlamaCppPythonProvider(llm)
 
 agent = LlamaCppAgent(
     provider,
-    system_prompt="You are Arsh, a helpful assistant.",
+    system_prompt="You are Arsh, a helpful assistant by Arshia Afshani. You should answer the user carefully.",
     predefined_messages_formatter_type=MessagesFormatterType.CHATML,
     debug_output=False
 )
 
-# 💬 تابع پاسخ‌دهنده
 def respond(message, chat_history, system_message, max_tokens, temperature, top_p, top_k, repeat_penalty):
-    # تنظیمات
     settings = provider.get_provider_default_settings()
     settings.temperature = temperature
     settings.top_k = top_k
@@ -44,7 +40,6 @@ def respond(message, chat_history, system_message, max_tokens, temperature, top_
     if chat_history is None:
         chat_history = []
 
-    # ✅ استفاده صحیح از BasicChatHistory
     messages = BasicChatHistory()
 
     for msg in chat_history:
@@ -53,7 +48,6 @@ def respond(message, chat_history, system_message, max_tokens, temperature, top_
         elif msg["role"] == "assistant":
             messages.add_message(AssistantMessage(msg["content"]))
 
-    # گرفتن استریم پاسخ
     stream = agent.get_chat_response(
         message,
         chat_history=messages,
@@ -67,15 +61,14 @@ def respond(message, chat_history, system_message, max_tokens, temperature, top_
         response += token
         yield response
 
-# 🎛️ رابط Gradio
 with gr.Blocks() as demo:
-    gr.Markdown("# Arsh-LLM Q4_K_M Chat Demo")
+    gr.Markdown("# Arsh-LLM Demo")
 
     with gr.Row():
         with gr.Column():
-            system_msg = gr.Textbox("You are Arsh, a helpful assistant.", label="System Message", interactive=True)
+            system_msg = gr.Textbox("You are Arsh, a helpful assistant by Arshia Afshani. You should answer the user carefully.", label="System Message", interactive=True)
             max_tokens = gr.Slider(1, 4096, value=2048, step=1, label="Max Tokens")
-            temperature = gr.Slider(0.1, 4.0, value=0.6, step=0.1, label="Temperature")
+            temperature = gr.Slider(0.1, 4.0, value=0.7, step=0.1, label="Temperature")
             top_p = gr.Slider(0.1, 1.0, value=0.95, step=0.05, label="Top-p")
             top_k = gr.Slider(0, 100, value=40, step=1, label="Top-k")
             repeat_penalty = gr.Slider(0.0, 2.0, value=1.1, step=0.1, label="Repetition Penalty")
@@ -98,6 +91,5 @@ with gr.Blocks() as demo:
         chatbot=chatbot
     )
 
-# 🚀 اجرای برنامه
 if __name__ == "__main__":
     demo.launch()