Update app.py

app.py

@@ -5,7 +5,6 @@ import warnings
 
 warnings.filterwarnings("ignore")
 
-# Global değişkenler
 model = None
 tokenizer = None
 model_loaded = False
@@ -17,16 +16,15 @@ def load_model():
         return True
     
     try:
-        print("[INFO] Model yükleniyor...")
+        print("[INFO] Model loading...")
         
         model_path = "Neurazum/Lbai-1-preview"
         
         tokenizer = AutoTokenizer.from_pretrained(model_path)
         
-        # CPU için optimize edilmiş yükleme
         model = AutoModelForCausalLM.from_pretrained(
             model_path,
-            torch_dtype=torch.float32,  # CPU için float32
+            torch_dtype=torch.float32,
             device_map="cpu",
             trust_remote_code=True,
             low_cpu_mem_usage=True
@@ -34,7 +32,7 @@ def load_model():
         
         model.eval()
         model_loaded = True
-        print("[INFO] Model başarıyla yüklendi!")
+        print("[INFO] Model successfully loaded!")
         return True
         
     except Exception as e:
@@ -45,33 +43,44 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
     global model, tokenizer, model_loaded
     
     if not message or message.strip() == "":
-        yield "Lütfen bir mesaj yazın."
+        yield "Please write a message..."
         return
     
-    # Model yüklü değilse yükle
     if not model_loaded:
-        yield "⏳ Model yükleniyor, lütfen bekleyin..."
+        yield "⏳ Loading model, please wait..."
         if not load_model():
-            yield "❌ Model yüklenemedi. Lütfen daha sonra tekrar deneyin."
+            yield "❌ Model could not be loaded. Please try again later."
             return
     
     try:
-        # Prompt oluştur
         prompt = f"{system_message}\n\n"
         
         if history:
-            for user_msg, assistant_msg in history:
-                if user_msg:
-                    prompt += f"Patient: {user_msg}\n"
-                if assistant_msg:
-                    prompt += f"Doctor: {assistant_msg}\n"
+            for item in history:
+                try:
+                    if isinstance(item, (list, tuple)) and len(item) >= 2:
+                        user_msg, assistant_msg = item[0], item[1]
+                        if user_msg:
+                            prompt += f"Patient: {user_msg}\n"
+                        if assistant_msg:
+                            prompt += f"Doctor: {assistant_msg}\n"
+                    elif isinstance(item, dict):
+                        role = item.get("role", "")
+                        content = item.get("content", "")
+                        if role == "user" and content:
+                            prompt += f"Patient: {content}\n"
+                        elif role == "assistant" and content:
+                            prompt += f"Doctor: {content}\n"
+                except Exception as e:
+                    print(f"[WARNING] History item skipped: {e}")
+                    continue
         
         prompt += f"Patient: {message}\nDoctor:"
         
-        # Tokenize
+        print(f"[DEBUG] Prompt: {prompt[:300]}...")
+        
         inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
         
-        # Generate
         with torch.no_grad():
             outputs = model.generate(
                 **inputs,
@@ -82,25 +91,27 @@ def respond(message, history, system_message, max_tokens, temperature, top_p):
                 pad_token_id=tokenizer.eos_token_id
             )
         
-        # Decode
         full_response = tokenizer.decode(outputs[0], skip_special_tokens=True)
         
-        # Sadece yeni üretilen kısmı al
         if "Doctor:" in full_response:
             response = full_response.split("Doctor:")[-1].strip()
         else:
             response = full_response[len(prompt):].strip()
         
-        yield response if response else "Model yanıt üretemedi."
+        if "\nPatient:" in response:
+            response = response.split("\nPatient:")[0].strip()
+        
+        yield response if response else "The model could not generate a response."
         
     except Exception as e:
+        import traceback
+        print(f"[ERROR] {traceback.format_exc()}")
         yield f"❌ Hata: {str(e)}"
 
-# Gradio arayüzü
 chatbot = gr.ChatInterface(
     fn=respond,
-    title="🩺 Medical AI Assistant",
-    description="Tıbbi sorularınızı sorun. İlk mesajda model yüklenecek, biraz bekleyin.",
+    title="Lbai-1-preview",
+    description="Ask your medical questions. The model will load in the first message, so please wait a moment. Artificial intelligence can make mistakes.",
     additional_inputs=[
         gr.Textbox(
             value="You are a helpful medical assistant.",