Update app.py

app.py

@@ -1,16 +1,14 @@
 import gradio as gr
 from huggingface_hub import hf_hub_download
-import os
 
-# Download model first
+# Download model
 print("📥 Downloading GGUF model...")
 model_path = hf_hub_download(
     repo_id="HexQuant/Pars-Medical-o1-Llama-FFT-GGUF",
-    filename="Pars-Medical-o1-Llama-FFT-Q4_K_M.gguf",
+    filename="model-Q4_K_M.gguf",  # ✅ Fixed filename
 )
 print(f"✅ Model downloaded: {model_path}")
 
-# Now import and load
 from llama_cpp import Llama
 
 print("🔄 Loading model...")
@@ -21,81 +19,63 @@ llm = Llama(
     n_gpu_layers=0,
     verbose=False,
 )
-print("✅ Model loaded!")
+print("✅ Model ready!")
 
-SYSTEM_PROMPT = """You are Pars-Medical, a bilingual (Persian/English) medical AI assistant.
+SYSTEM = """You are Pars-Medical, a bilingual (Persian/English) medical AI assistant with chain-of-thought reasoning.
+When answering: analyze carefully, think step-by-step, provide accurate medical information.
+Respond in the user's language. ⚠️ Educational purposes only - always consult a real doctor."""
 
-When answering medical questions:
-1. Analyze the question carefully
-2. Think step-by-step (you can use <think> tags for reasoning)
-3. Provide accurate, helpful medical information
 
-You respond in the same language as the user (Persian or English).
-
-⚠️ Disclaimer: This is for educational purposes only. Always consult a real doctor."""
-
-
-def respond(message, history):
-    messages = [{"role": "system", "content": SYSTEM_PROMPT}]
-    
-    for user_msg, assistant_msg in history:
-        messages.append({"role": "user", "content": user_msg})
-        if assistant_msg:
-            messages.append({"role": "assistant", "content": assistant_msg})
-    
+def chat(message, history):
+    messages = [{"role": "system", "content": SYSTEM}]
+    for user, assistant in history:
+        messages.append({"role": "user", "content": user})
+        if assistant:
+            messages.append({"role": "assistant", "content": assistant})
     messages.append({"role": "user", "content": message})
     
-    response = llm.create_chat_completion(
-        messages=messages,
-        max_tokens=512,
-        temperature=0.7,
-        stream=True,
-    )
+    response = llm.create_chat_completion(messages=messages, max_tokens=512, temperature=0.7, stream=True)
     
-    partial = ""
+    text = ""
     for chunk in response:
-        if chunk["choices"][0].get("delta", {}).get("content"):
-            partial += chunk["choices"][0]["delta"]["content"]
-            yield partial
-
+        delta = chunk["choices"][0].get("delta", {}).get("content", "")
+        text += delta
+        yield text
 
-EXAMPLES = [
-    ["علائم دیابت نوع ۲ چیست؟"],
-    ["What are the symptoms of hypothyroidism?"],
-    ["سردرد میگرنی چه علائمی دارد؟"],
-    ["How to differentiate viral from bacterial infection?"],
-]
 
-with gr.Blocks(title="🩺 Pars-Medical-o1") as demo:
+with gr.Blocks(title="🩺 Pars-Medical-o1", theme=gr.themes.Soft()) as demo:
     gr.HTML("""
     <div style="text-align:center; padding:20px;">
         <h1>🩺 Pars-Medical-o1</h1>
-        <p>دستیار پزشکی دوزبانه فارسی-انگلیسی | Bilingual Medical AI</p>
-        <p>
-            <a href="https://huggingface.co/HexQuant/Pars-Medical-o1-Llama-FFT">🤗 Model</a> |
-            <a href="https://twitter.com/xPOURY4">🐦 @xPOURY4</a>
-        </p>
+        <h3>دستیار پزشکی هوشمند دوزبانه | Bilingual Medical AI Assistant</h3>
+        <p><a href="https://huggingface.co/HexQuant/Pars-Medical-o1-Llama-FFT" target="_blank">🤗 Model</a> • <a href="https://twitter.com/xPOURY4" target="_blank">🐦 @xPOURY4</a> • <a href="https://github.com/xPOURY4" target="_blank">💻 GitHub</a></p>
     </div>
     """)
     
-    chatbot = gr.Chatbot(height=400)
-    msg = gr.Textbox(placeholder="سوال پزشکی خود را بپرسید... | Ask your question...", show_label=False)
+    chatbot = gr.Chatbot(height=450, placeholder="سوال پزشکی بپرسید... | Ask a medical question...")
+    msg = gr.Textbox(placeholder="پیام خود را بنویسید... | Type your message...", show_label=False)
     
     with gr.Row():
-        submit = gr.Button("ارسال | Send", variant="primary")
-        clear = gr.Button("پاک کردن | Clear")
+        send = gr.Button("📤 ارسال | Send", variant="primary")
+        clear = gr.Button("🗑️ پاک کردن | Clear")
     
-    gr.Examples(examples=EXAMPLES, inputs=msg)
+    gr.Examples([
+        ["علائم دیابت نوع ۲ چیست؟"],
+        ["What are symptoms of hypothyroidism?"],
+        ["تفاوت میگرن و سردرد تنشی چیست؟"],
+        ["How to differentiate viral vs bacterial infection?"],
+    ], inputs=msg, label="💡 نمونه سوالات | Examples")
     
     gr.HTML("""
-    <div style="text-align:center; padding:10px; background:#fff3cd; border-radius:8px; margin-top:15px;">
-        ⚠️ فقط برای آموزش | Educational purposes only - Not medical advice
+    <div style="background:#fff3cd; padding:12px; border-radius:8px; text-align:center; margin-top:15px;">
+        ⚠️ <b>توجه:</b> فقط برای آموزش است، جایگزین مشاوره پزشکی نیست<br>
+        <b>Disclaimer:</b> Educational only, not medical advice
     </div>
-    <p style="text-align:center; color:#888; margin-top:10px;">Made with ❤️ by @xPOURY4 🇮🇷</p>
+    <p style="text-align:center; color:#666; margin-top:15px;">Made with ❤️ by <a href="https://twitter.com/xPOURY4">@xPOURY4</a> in Iran 🇮🇷</p>
     """)
     
-    msg.submit(respond, [msg, chatbot], chatbot).then(lambda: "", None, msg)
-    submit.click(respond, [msg, chatbot], chatbot).then(lambda: "", None, msg)
+    msg.submit(chat, [msg, chatbot], chatbot).then(lambda: "", None, msg)
+    send.click(chat, [msg, chatbot], chatbot).then(lambda: "", None, msg)
     clear.click(lambda: None, None, chatbot)
 
 demo.queue().launch()