added prompt for model behaviour

app.py

@@ -112,15 +112,22 @@ def generate_response(prompt, max_tokens=None, temperature=None, top_p=None):
     print(f"📊 Generation params: max_tokens={max_tokens}, temp={temperature}, top_p={top_p}")
     
     try:
-        # Use the medical system prompt
-        full_prompt = f"{MEDICAL_SYSTEM_PROMPT}\n\nPatient/User: {prompt}\nMedical Assistant:"
+        # Format prompt for DialoGPT - simpler approach
+        if "dialogpt" in current_model_name.lower():
+            # DialoGPT works better with direct conversation format
+            full_prompt = f"Medical Question: {prompt}\nMedical Assistant:"
+        else:
+            # Use full system prompt for other models
+            full_prompt = f"{MEDICAL_SYSTEM_PROMPT}\n\nPatient/User: {prompt}\nMedical Assistant:"
+        
+        print(f"📝 Using prompt format: {full_prompt[:100]}{'...' if len(full_prompt) > 100 else ''}")
         
         # Tokenize input with proper truncation
         inputs = tokenizer(
             full_prompt, 
             return_tensors="pt", 
             truncation=True, 
-            max_length=1024,
+            max_length=512,  # Reduced for DialoGPT
             padding=True
         )
         
@@ -162,11 +169,12 @@ def generate_response(prompt, max_tokens=None, temperature=None, top_p=None):
             # Fallback extraction
             response = full_response[len(full_prompt):].strip()
         
-        # Clean up response
-        if not response or len(response.strip()) < 10:
-            response = "I understand you're asking about a medical topic. While I'd like to help, I recommend consulting with a qualified healthcare professional who can provide personalized advice based on your specific situation."
+        # Clean up response - keep it natural as per CareConnect guidelines
+        if not response or len(response.strip()) < 3:
+            response = "Could you please provide more details about your question so I can help you better? 😊"
         
         print(f"✅ Generated response length: {len(response)} characters")
+        print(f"📄 Response preview: {response[:150]}{'...' if len(response) > 150 else ''}")
         
         # Clean up memory
         del inputs, outputs

app_optimized.py

@@ -0,0 +1,196 @@
+import gradio as gr
+import torch
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import logging
+import gc
+import warnings
+import os
+from huggingface_hub import login
+
+# Login with the secret token
+login(token=os.getenv("HF_TOKEN"))
+
+# Suppress warnings
+warnings.filterwarnings("ignore")
+logging.getLogger("transformers").setLevel(logging.ERROR)
+
+# Configuration for optimized performance
+MODEL_NAME = "microsoft/DialoGPT-medium"
+MAX_NEW_TOKENS = 150
+TEMPERATURE = 0.8
+TOP_P = 0.9
+
+# Medical enhancement prompt - detailed CareConnect specifications
+MEDICAL_CONTEXT = """You are a friendly and smart medical assistant. Your job is to give short, clear, and helpful health information.
+      
+Your answers should:
+- Stay focused. No long essays or extra fluff.
+- Give basic helpful steps for common symptoms like fever, cough, or headache (e.g., rest, drink fluids, take paracetamol if needed).
+- For any serious or unclear issues, remind the user to see a doctor — but do it briefly and naturally.
+- Keep responses concise and under 4 sentences when possible.
+      
+Tone:
+- Friendly, supportive, and calm.
+- No robotic warnings unless needed. Keep it real and human.
+- Use emojis like 😊 or 👍 occasionally to appear friendly.
+      
+Important rules:
+- NEVER include text in parentheses in your responses.
+- NEVER include any meta-instructions in your responses.
+- NEVER include reminders about what you should do in future responses.
+- DO NOT include phrases like "We're here to help" or "I'm just an AI".
+- DO NOT include any text that instructs you what to do or how to behave.
+- DO NOT include any sentences that start with "If the user asks..." or "Remember..."
+- DO NOT include "(smile)" - instead, use actual emojis like 😊 or 👍 when appropriate.
+- DO NOT include numbered references like [1], [2], etc. in your responses.
+- DO NOT include any text that explains what your response is doing.
+- DO NOT include "user:" or "assistant:" prefixes in your responses.
+- DO NOT include hypothetical user questions in your responses.
+- DO NOT refuse to answer harmless non-medical questions like jokes or general knowledge.
+- Don't give exact dosages or diagnoses.
+- Be consistent in your responses regardless of the user's role."""
+
+# Global variables
+model = None
+tokenizer = None
+
+def load_model():
+    """Load DialoGPT model optimized for CPU"""
+    global model, tokenizer
+    
+    try:
+        print(f"🏥 Loading medical chatbot model: {MODEL_NAME}")
+        
+        # Load tokenizer
+        tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, padding_side="left")
+        if tokenizer.pad_token is None:
+            tokenizer.pad_token = tokenizer.eos_token
+        
+        # Load model with CPU optimization
+        model = AutoModelForCausalLM.from_pretrained(
+            MODEL_NAME,
+            torch_dtype=torch.float32,  # Use float32 for CPU
+            low_cpu_mem_usage=True,
+            trust_remote_code=True
+        )
+        
+        print(f"✅ Model loaded successfully!")
+        return True
+        
+    except Exception as e:
+        print(f"❌ Failed to load model: {str(e)}")
+        return False
+
+def generate_medical_response(prompt):
+    """Generate medical response with DialoGPT"""
+    global model, tokenizer
+    
+    if model is None or tokenizer is None:
+        return "❌ Model not loaded. Please wait for initialization."
+    
+    try:
+        # Enhanced prompt for medical context
+        medical_prompt = f"{MEDICAL_CONTEXT}\n\nUser: {prompt}\nAssistant:"
+        
+        print(f"🔄 Processing: {prompt[:50]}{'...' if len(prompt) > 50 else ''}")
+        
+        # Tokenize
+        inputs = tokenizer.encode(medical_prompt, return_tensors="pt", max_length=400, truncation=True)
+        
+        # Generate with optimized parameters
+        with torch.no_grad():
+            outputs = model.generate(
+                inputs,
+                max_new_tokens=MAX_NEW_TOKENS,
+                temperature=TEMPERATURE,
+                top_p=TOP_P,
+                do_sample=True,
+                pad_token_id=tokenizer.eos_token_id,
+                repetition_penalty=1.1,
+                early_stopping=True,
+                num_return_sequences=1
+            )
+        
+        # Decode response
+        full_response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+        
+        # Extract only the assistant's response
+        if "Assistant:" in full_response:
+            response = full_response.split("Assistant:")[-1].strip()
+        else:
+            response = full_response[len(medical_prompt):].strip()
+        
+        # Clean up response - keep it natural as per prompt guidelines
+        if not response or len(response) < 10:
+            response = "I'd be happy to help with your medical question. Could you please provide more specific details? 😊"
+        
+        print(f"✅ Response generated: {len(response)} characters")
+        
+        # Memory cleanup
+        del inputs, outputs
+        gc.collect()
+        
+        return response
+        
+    except Exception as e:
+        print(f"❌ Generation error: {str(e)}")
+        return f"I encountered a technical issue. Please try rephrasing your question. For immediate medical concerns, please consult a healthcare professional."
+
+def chat_interface(message, history):
+    """Main chat interface function"""
+    if not message or not message.strip():
+        return "Please enter a medical question."
+    
+    # Generate response
+    response = generate_medical_response(message.strip())
+    
+    return response
+
+# Load model on startup
+print("🏥 Initializing Medical Chatbot...")
+model_loaded = load_model()
+
+if not model_loaded:
+    print("⚠️ WARNING: Model failed to load. Responses may be limited.")
+
+# Create Gradio interface
+demo = gr.ChatInterface(
+    chat_interface,
+    type="messages",
+    title="🏥 Medical Information Assistant",
+    description="""
+    A medical information chatbot powered by AI. This assistant provides educational health information.
+    
+    ⚠️ **Important Disclaimer**: This chatbot provides general health information for educational purposes only. 
+    It should not replace professional medical advice, diagnosis, or treatment. Always consult qualified 
+    healthcare professionals for medical concerns.
+    """,
+    examples=[
+        "What are the symptoms of diabetes?",
+        "How can I maintain a healthy heart?", 
+        "What should I know about high blood pressure?",
+        "Tell me about the importance of regular exercise",
+        "What are common causes of headaches?",
+        "How can I improve my sleep quality?"
+    ],
+    cache_examples=False,
+    theme=gr.themes.Soft(),
+    css="""
+    .gradio-container {
+        max-width: 800px !important;
+        margin: auto !important;
+    }
+    .message {
+        border-radius: 10px !important;
+    }
+    """
+)
+
+if __name__ == "__main__":
+    demo.launch(
+        server_name="0.0.0.0",
+        server_port=7860,
+        share=True,
+        show_error=True,
+        debug=True
+    )

config.py

@@ -54,17 +54,35 @@ GENERATION_DEFAULTS = {
     "no_repeat_ngram_size": 2   # Reduced for better performance
 }
 
-# Enhanced medical prompt template optimized for DialoGPT
-MEDICAL_SYSTEM_PROMPT = """You are a knowledgeable medical information assistant. Your responses should be:
-
-🩺 MEDICAL FOCUS: Provide evidence-based health information from reliable medical sources
-⚠️ SAFETY FIRST: Always recommend consulting healthcare professionals for diagnosis and treatment
-📚 EDUCATIONAL: Explain medical concepts clearly for general understanding
-🤝 SUPPORTIVE: Be empathetic while maintaining professional boundaries
-
-REMINDER: This is educational information only - not medical advice. Always consult healthcare professionals for personal medical concerns.
-
-Respond as a helpful medical information assistant would, focusing on accurate, educational health content."""
+# Medical system prompt from CareConnect - detailed specifications
+MEDICAL_SYSTEM_PROMPT = """You are a friendly and smart medical assistant. Your job is to give short, clear, and helpful health information.
+      
+Your answers should:
+- Stay focused. No long essays or extra fluff.
+- Give basic helpful steps for common symptoms like fever, cough, or headache (e.g., rest, drink fluids, take paracetamol if needed).
+- For any serious or unclear issues, remind the user to see a doctor — but do it briefly and naturally.
+- Keep responses concise and under 4 sentences when possible.
+      
+Tone:
+- Friendly, supportive, and calm.
+- No robotic warnings unless needed. Keep it real and human.
+- Use emojis like 😊 or 👍 occasionally to appear friendly.
+      
+Important rules:
+- NEVER include text in parentheses in your responses.
+- NEVER include any meta-instructions in your responses.
+- NEVER include reminders about what you should do in future responses.
+- DO NOT include phrases like "We're here to help" or "I'm just an AI".
+- DO NOT include any text that instructs you what to do or how to behave.
+- DO NOT include any sentences that start with "If the user asks..." or "Remember..."
+- DO NOT include "(smile)" - instead, use actual emojis like 😊 or 👍 when appropriate.
+- DO NOT include numbered references like [1], [2], etc. in your responses.
+- DO NOT include any text that explains what your response is doing.
+- DO NOT include "user:" or "assistant:" prefixes in your responses.
+- DO NOT include hypothetical user questions in your responses.
+- DO NOT refuse to answer harmless non-medical questions like jokes or general knowledge.
+- Don't give exact dosages or diagnoses.
+- Be consistent in your responses regardless of the user's role."""
 
 # UI settings
 UI_CONFIG = {