Update app.py

app.py

@@ -3,11 +3,11 @@ import torch
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
 import traceback
 
-MODEL_NAME = "Qwen/Qwen2.5-3B-Instruct"
+# 👈 استخدم موديلك انت
+MODEL_NAME = "kawkabelaloom/astramindx"
 
 SYSTEM_PROMPT = "أنت مساعد عربي ذكي، تجيب بوضوح وبأسلوب بسيط ومفيد."
 
-# ---------- تحميل الموديل ----------
 print("🔄 Loading tokenizer...")
 tokenizer = AutoTokenizer.from_pretrained(
     MODEL_NAME,
@@ -17,8 +17,9 @@ tokenizer = AutoTokenizer.from_pretrained(
 print("🔄 Loading model (CPU)...")
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_NAME,
-    torch_dtype=torch.float32,
     device_map="cpu",
+    torch_dtype=torch.float32,
+    low_cpu_mem_usage=True,
     trust_remote_code=True
 )
 
@@ -30,54 +31,37 @@ generator = pipeline(
     tokenizer=tokenizer
 )
 
-# ---------- تنسيق البرومبت ----------
 def build_prompt(history, user_message):
     prompt = SYSTEM_PROMPT + "\n\n"
     for user, bot in history:
-        prompt += f"المستخدم: {user}\n"
-        prompt += f"المساعد: {bot}\n"
+        prompt += f"المستخدم: {user}\nالمساعد: {bot}\n"
     prompt += f"المستخدم: {user_message}\nالمساعد:"
     return prompt
 
-
-# ---------- دالة الرد ----------
 def chat(user_message, history):
-    if not user_message.strip():
-        return history, "⚠️ اكتب رسالة"
-
     try:
         prompt = build_prompt(history, user_message)
 
         output = generator(
             prompt,
-            max_new_tokens=256,
+            max_new_tokens=128,  # 👈 قللها للسرعة
             temperature=0.7,
             top_p=0.9,
             do_sample=True
         )
 
-        full_text = output[0]["generated_text"]
-        response = full_text.split("المساعد:")[-1].strip()
-
+        response = output[0]["generated_text"].split("المساعد:")[-1].strip()
         history.append((user_message, response))
         return history, ""
 
     except Exception as e:
-        error_msg = f"❌ Error:\n{str(e)}\n\n{traceback.format_exc()}"
-        history.append((user_message, error_msg))
+        history.append((user_message, str(e)))
         return history, ""
 
-
-# ---------- واجهة Gradio ----------
-with gr.Blocks(title="Qwen2.5 Arabic Chatbot") as demo:
-    gr.Markdown("# 🤖 Qwen2.5 Arabic Chatbot")
-    gr.Markdown("نموذج Qwen2.5 يعمل بدون messages وبدون API")
-
+with gr.Blocks() as demo:
+    gr.Markdown("# 🤖 Astramindx Chatbot")
     chatbot = gr.Chatbot(height=450)
-    msg = gr.Textbox(label="رسالتك", placeholder="اكتب سؤالك هنا...")
-    clear = gr.Button("🗑️ مسح المحادثة")
-
+    msg = gr.Textbox()
     msg.submit(chat, [msg, chatbot], [chatbot, msg])
-    clear.click(lambda: [], None, chatbot)
 
 demo.launch()