Delete medical_chatbot.py

medical_chatbot.py

@@ -1,866 +0,0 @@
-
-# Setup and Installation
-
-import torch
-print("🖥️ System Check:")
-print(f"CUDA available: {torch.cuda.is_available()}")
-if torch.cuda.is_available():
-    print(f"GPU device: {torch.cuda.get_device_name(0)}")
-    print(f"GPU memory: {torch.cuda.get_device_properties(0).total_memory / 1e9:.1f} GB")
-else:
-    print("⚠️ No GPU detected - BioGPT will run on CPU (much slower)")
-
-print("\n🔧 Installing required packages...")
-
-# Import Libraries
-
-import os
-import re
-import torch
-import warnings
-import numpy as np
-import faiss  # FAISS for vector search
-from transformers import (
-    AutoTokenizer,
-    AutoModelForCausalLM,
-    pipeline,
-    BitsAndBytesConfig
-)
-from sentence_transformers import SentenceTransformer
-from typing import List, Dict, Optional
-import time
-from datetime import datetime
-import json
-import pickle
-
-# Suppress warnings for cleaner output
-warnings.filterwarnings('ignore')
-
-print("📚 Libraries imported successfully!")
-print(f"🔍 FAISS version: {faiss.__version__}")
-print("🎯 Using FAISS for vector search (ChromaDB completely removed)")
-
-# File Upload Helper
-
-from google.colab import files
-import io
-
-def upload_medical_data():
-    """Upload your Pediatric_cleaned.txt file"""
-    print("📁 Please upload your Pediatric_cleaned.txt file:")
-    uploaded = files.upload()
-
-    # Get the uploaded file
-    filename = list(uploaded.keys())[0]
-    print(f"✅ File '{filename}' uploaded successfully!")
-
-    # Read the content
-    content = uploaded[filename].decode('utf-8')
-
-    # Save it locally in Colab
-    with open('Pediatric_cleaned.txt', 'w', encoding='utf-8') as f:
-        f.write(content)
-
-    print(f"📝 File saved as 'Pediatric_cleaned.txt' ({len(content)} characters)")
-    return 'Pediatric_cleaned.txt'
-
-medical_file = 'Pediatric_cleaned.txt'
-
-# BioGPT Medical Chatbot Class
-
-from typing import List, Dict, Optional # Import List, Dict, Optional
-
-class ColabBioGPTChatbot:
-    def setup_biogpt(self):
-        """Setup BioGPT model with fallback to base BioGPT if Large fails"""
-        print("🧠 Attempting to load BioGPT-Large...")
-        model_name = "microsoft/BioGPT-Large"
-
-        try:
-            if self.use_8bit:
-                quantization_config = BitsAndBytesConfig(
-                    load_in_8bit=True,
-                    llm_int8_threshold=6.0,
-                    llm_int8_has_fp16_weight=False,
-                )
-            else:
-                quantization_config = None
-
-            self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-            if self.tokenizer.pad_token is None:
-                self.tokenizer.pad_token = self.tokenizer.eos_token
-
-            self.model = AutoModelForCausalLM.from_pretrained(
-                model_name,
-                quantization_config=quantization_config,
-                torch_dtype=torch.float16 if self.device == "cuda" else torch.float32,
-                device_map="auto" if self.device == "cuda" else None,
-                trust_remote_code=True
-            )
-
-            if self.device == "cuda" and quantization_config is None:
-                self.model = self.model.to(self.device)
-
-            print("✅ BioGPT-Large loaded successfully!")
-            self.test_biogpt()
-
-        except Exception as e:
-            print(f"❌ BioGPT-Large loading failed: {e}")
-            print("🔁 Falling back to base model: microsoft/BioGPT")
-            self.setup_fallback_biogpt()
-
-        def setup_fallback_biogpt(self):
-        """Fallback to microsoft/BioGPT if BioGPT-Large fails"""
-        model_name = "microsoft/BioGPT"
-
-        try:
-            self.tokenizer = AutoTokenizer.from_pretrained(model_name)
-            if self.tokenizer.pad_token is None:
-                self.tokenizer.pad_token = self.tokenizer.eos_token
-
-            self.model = AutoModelForCausalLM.from_pretrained(
-                model_name,
-                torch_dtype=torch.float32,
-                trust_remote_code=True
-            )
-            self.model = self.model.to(self.device)
-
-            print("✅ Base BioGPT model loaded as fallback.")
-            self.test_biogpt()
-
-        except Exception as e:
-            print(f"❌ Failed to load fallback BioGPT: {e}")
-            self.model = None
-            self.tokenizer = None
-
-(self):
-        if self.tokenizer is None or self.model is None:
-            print("⚠️ Skipping test — model or tokenizer not available.")
-            return
-        print("🧪 Testing BioGPT...")
-        try:
-            input_text = "What is pneumonia?"
-            inputs = self.tokenizer(input_text, return_tensors="pt").to(self.device)
-            outputs = self.model.generate(**inputs, max_new_tokens=30)
-            response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
-            print(f"🧠 Model output: {response}")
-        except Exception as e:
-            print(f"⚠️ BioGPT test failed: {e}")
-
-        def __init__(self, use_gpu=True, use_8bit=True):
-        """Initialize BioGPT chatbot optimized for Google Colab"""
-        print("🏥 Initializing Professional BioGPT Medical Chatbot...")
-
-        self.device = "cuda" if torch.cuda.is_available() and use_gpu else "cpu"
-        self.use_8bit = use_8bit and torch.cuda.is_available()
-
-        print(f"🖥️ Using device: {self.device}")
-        if self.use_8bit:
-            print("💾 Using 8-bit quantization for memory efficiency")
-
-        # Setup components
-        self.setup_embeddings()
-        self.setup_faiss_index() # Ensure this sets up self.collection if needed
-        self.setup_biogpt()
-
-        # Conversation tracking
-        self.conversation_history = []
-        self.knowledge_chunks = []
-
-        print("✅ BioGPT Medical Chatbot ready for professional medical assistance!")
-
-    def setup_embeddings(self):
-        """Setup medical-optimized embeddings"""
-        print("🔧 Loading medical embeddings...")
-        try:
-            # Use a medical-focused embedding model if available, otherwise general
-            self.embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
-            self.embedding_dim = self.embedding_model.get_sentence_embedding_dimension()
-            print(f"✅ Embeddings loaded (dimension: {self.embedding_dim})")
-            self.use_embeddings = True
-        except Exception as e:
-            print(f"⚠️ Embeddings failed: {e}")
-            self.embedding_model = None
-            self.embedding_dim = 384 # default dimension
-            self.use_embeddings = False
-
-    def setup_faiss_index(self):
-        """Setup faiss for CPU-based vector search"""
-        print("🔧 Setting up FAISS vector database...")
-        try:
-            print(' Using CPU FAISS index for maximum compatibility')
-            self.faiss_index = faiss.IndexFlatIP(self.embedding_dim)  # In-memory for Colab
-            self.use_gpu_faiss = False
-            self.faiss_ready = True # Set to True when index is ready
-            self.collection = self.faiss_index # Initialize collection attribute
-            print("✅ FAISS CPU index initialized successfully")
-        except Exception as e:
-            print(f"❌ FAISS setup failed: {e}")
-            self.faiss_index = None
-            self.faiss_ready = False
-            self.collection = None # Ensure collection is None on failure
-
-    
-        """Test BioGPT with a simple medical query"""
-        print("🧪 Testing BioGPT...")
-        if not self.model or not self.tokenizer:
-            print("⚠️ BioGPT model or tokenizer not loaded. Skipping test.")
-            return
-        try:
-            test_prompt = "Fever in children can be caused by"
-            inputs = self.tokenizer(test_prompt, return_tensors="pt")
-
-            if self.device == "cuda": # Ensure inputs are on the correct device
-                inputs = {k: v.to(self.device) for k, v in inputs.items()}
-
-            with torch.no_grad():
-                outputs = self.model.generate(
-                    **inputs,
-                    max_new_tokens=20,
-                    do_sample=True,
-                    temperature=0.7,
-                    pad_token_id=self.tokenizer.eos_token_id
-                )
-
-            response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
-            print(f"✅ BioGPT test successful!")
-            print(f"   Test response: {response}")
-
-        except Exception as e:
-            print(f"⚠️ BioGPT test failed: {e}")
-
-    def load_medical_data(self, file_path):
-    """Load and process medical data with progress tracking"""
-        print(f"📖 Loading medical data from {file_path}...")
-
-        try:
-            with open(file_path, 'r', encoding='utf-8') as f:
-                text = f.read()
-            print(f"📄 File loaded: {len(text):,} characters")
-        except FileNotFoundError:
-            print(f"❌ File {file_path} not found!")
-            raise ValueError("Failed to load medical data")
-
-        # Create chunks optimized for medical content
-        print("📝 Creating medical-optimized chunks...")
-        chunks = self.create_medical_chunks(text)
-        print(f"📋 Created {len(chunks)} medical chunks")
-
-        self.knowledge_chunks = chunks
-
-        # Generate embeddings with progress and add to FAISS index
-        if self.use_embeddings and self.embedding_model and self.faiss_ready:
-            return self.generate_embeddings_with_progress(chunks)
-
-        print("✅ Medical data loaded (text search mode)")
-        return
-
-    def create_medical_chunks(self, text: str, chunk_size: int = 400) -> List[Dict]:
-        """Create medically-optimized text chunks"""
-        chunks = []
-
-        # Split by medical sections first
-        medical_sections = self.split_by_medical_sections(text)
-
-        chunk_id = 0
-        for section in medical_sections:
-            if len(section.split()) > chunk_size:
-                # Split large sections by sentences
-                sentences = re.split(r'[.!?]+', section)
-                current_chunk = ""
-
-                for sentence in sentences:
-                    sentence = sentence.strip()
-                    if not sentence:
-                        continue
-
-                    if len(current_chunk.split()) + len(sentence.split()) < chunk_size:
-                        current_chunk += sentence + ". "
-                    else:
-                        if current_chunk.strip():
-                            chunks.append({
-                                'id': chunk_id,
-                                'text': current_chunk.strip(),
-                                'medical_focus': self.identify_medical_focus(current_chunk)
-                            })
-                            chunk_id += 1
-                        current_chunk = sentence + ". "
-
-                if current_chunk.strip():
-                    chunks.append({
-                        'id': chunk_id,
-                        'text': current_chunk.strip(),
-                        'medical_focus': self.identify_medical_focus(current_chunk)
-                    })
-                    chunk_id += 1
-            else:
-                chunks.append({
-                    'id': chunk_id,
-                    'text': section,
-                    'medical_focus': self.identify_medical_focus(section)
-                })
-                chunk_id += 1
-
-        return chunks
-
-    def split_by_medical_sections(self, text: str) -> List[str]:
-        """Split text by medical sections"""
-        # Look for medical section headers
-        section_patterns = [
-            r'\n\s*(?:SYMPTOMS?|TREATMENT|DIAGNOSIS|CAUSES?|PREVENTION|MANAGEMENT).*?\n',
-            r'\n\s*\d+\.\s+',  # Numbered sections
-            r'\n\n+'  # Paragraph breaks
-        ]
-
-        sections = [text]
-        for pattern in section_patterns:
-            new_sections = []
-            for section in sections:
-                splits = re.split(pattern, section, flags=re.IGNORECASE)
-                new_sections.extend([s.strip() for s in splits if len(s.strip()) > 100])
-            sections = new_sections
-
-        return sections
-
-    def identify_medical_focus(self, text: str) -> str:
-        """Identify the medical focus of a text chunk"""
-        text_lower = text.lower()
-
-        # Medical categories
-        categories = {
-            'pediatric_symptoms': ['fever', 'cough', 'rash', 'vomiting', 'diarrhea'],
-            'treatments': ['treatment', 'therapy', 'medication', 'antibiotics'],
-            'diagnosis': ['diagnosis', 'diagnostic', 'symptoms', 'signs'],
-            'emergency': ['emergency', 'urgent', 'serious', 'hospital'],
-            'prevention': ['prevention', 'vaccine', 'immunization', 'avoid']
-        }
-
-        for category, keywords in categories.items():
-            if any(keyword in text_lower for keyword in keywords):
-                return category
-
-        return 'general_medical'
-
-    def generate_embeddings_with_progress(self, chunks: List[Dict]) -> bool:
-        """Generate embeddings with progress tracking and add to FAISS index"""
-        print("🔮 Generating medical embeddings and adding to FAISS index...")
-
-        if not self.embedding_model or not self.faiss_index:
-            print("❌ Embedding model or FAISS index not available.")
-            raise ValueError("Failed to load medical data")
-
-        try:
-            texts = [chunk['text'] for chunk in chunks]
-
-            # Generate embeddings in batches with progress
-            batch_size = 32
-            all_embeddings = []
-
-            for i in range(0, len(texts), batch_size):
-                batch_texts = texts[i:i+batch_size]
-                batch_embeddings = self.embedding_model.encode(batch_texts, show_progress_bar=False)
-                all_embeddings.extend(batch_embeddings)
-
-                # Show progress
-                progress = min(i + batch_size, len(texts))
-                print(f"   Progress: {progress}/{len(texts)} chunks processed", end='\r')
-
-            print(f"\n   ✅ Generated embeddings for {len(texts)} chunks")
-
-            # Add embeddings to FAISS index
-            print("💾 Adding embeddings to FAISS index...")
-            self.faiss_index.add(np.array(all_embeddings))
-
-            print("✅ Medical embeddings added to FAISS index successfully!")
-            return True
-
-        except Exception as e:
-            print(f"❌ Embedding generation or FAISS add failed: {e}")
-            raise ValueError("Failed to load medical data")
-
-
-    def retrieve_medical_context(self, query: str, n_results: int = 3) -> List[str]:
-        """Retrieve relevant medical context using embeddings or keyword search"""
-        if self.use_embeddings and self.embedding_model and self.faiss_ready:
-            try:
-                # Generate query embedding
-                query_embedding = self.embedding_model.encode([query])
-
-                # Search for similar content in FAISS index
-                distances, indices = self.faiss_index.search(np.array(query_embedding), n_results)
-
-                # Retrieve the corresponding chunks
-                context_chunks = [self.knowledge_chunks[i]['text'] for i in indices[0] if i != -1]
-
-                if context_chunks:
-                    return context_chunks
-
-            except Exception as e:
-                print(f"⚠️ Embedding search failed: {e}")
-
-        # Fallback to keyword search
-        print("⚠️ Falling back to keyword search.")
-        return self.keyword_search_medical(query, n_results)
-
-
-    def keyword_search_medical(self, query: str, n_results: int) -> List[str]:
-        """Medical-focused keyword search"""
-        if not self.knowledge_chunks:
-            return []
-
-        query_words = set(query.lower().split())
-        chunk_scores = []
-
-        for chunk_info in self.knowledge_chunks:
-            chunk_text = chunk_info['text']
-            chunk_words = set(chunk_text.lower().split())
-
-            # Calculate relevance score
-            word_overlap = len(query_words.intersection(chunk_words))
-            base_score = word_overlap / len(query_words) if query_words else 0
-
-            # Boost medical content
-            medical_boost = 0
-            if chunk_info.get('medical_focus') in ['pediatric_symptoms', 'treatments', 'diagnosis']:
-                medical_boost = 0.5
-
-            final_score = base_score + medical_boost
-
-            if final_score > 0:
-                chunk_scores.append((final_score, chunk_text))
-
-        # Return top matches
-        chunk_scores.sort(reverse=True)
-        return [chunk for _, chunk in chunk_scores[:n_results]]
-
-    def generate_biogpt_response(self, context: str, query: str) -> str:
-        """Generate medical response using BioGPT"""
-        if not self.model or not self.tokenizer:
-            return "Medical model not available. Please check the setup."
-
-        try:
-            # Create medical-focused prompt
-            prompt = f"""Medical Context: {context[:800]}
-
-Question: {query}
-
-Medical Answer:"""
-
-            # Tokenize input
-            inputs = self.tokenizer(
-                prompt,
-                return_tensors="pt",
-                truncation=True,
-                max_length=1024
-            )
-
-            # Move inputs to the correct device
-            if self.device == "cuda":
-                inputs = {k: v.to(self.device) for k, v in inputs.items()}
-
-            # Generate response
-            with torch.no_grad():
-                outputs = self.model.generate(
-                    **inputs,
-                    max_new_tokens=150,
-                    do_sample=True,
-                    temperature=0.7,
-                    top_p=0.9,
-                    pad_token_id=self.tokenizer.eos_token_id,
-                    repetition_penalty=1.1
-                )
-
-            # Decode response
-            full_response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-            # Extract just the generated part
-            if "Medical Answer:" in full_response:
-                generated_response = full_response.split("Medical Answer:")[-1].strip()
-            else:
-                generated_response = full_response[len(prompt):].strip()
-
-            # Clean up response
-            cleaned_response = self.clean_medical_response(generated_response)
-
-            return cleaned_response
-
-        except Exception as e:
-            print(f"⚠️ BioGPT generation failed: {e}")
-            return self.fallback_response(context, query)
-
-    def clean_medical_response(self, response: str) -> str:
-        """Clean and format medical response"""
-        # Remove incomplete sentences and limit length
-        sentences = re.split(r'[.!?]+', response)
-        clean_sentences = []
-
-        for sentence in sentences:
-            sentence = sentence.strip()
-            if len(sentence) > 10 and not sentence.endswith(('and', 'or', 'but', 'however')):
-                clean_sentences.append(sentence)
-            if len(clean_sentences) >= 3:  # Limit to 3 sentences
-                break
-
-        if clean_sentences:
-            cleaned = '. '.join(clean_sentences) + '.'
-        else:
-            cleaned = response[:200] + '...' if len(response) > 200 else response
-
-        return cleaned
-
-    def fallback_response(self, context: str, query: str) -> str:
-        """Fallback response when BioGPT fails"""
-        # Extract key sentences from context
-        sentences = [s.strip() for s in context.split('.') if len(s.strip()) > 20]
-
-        if sentences:
-            response = sentences[0] + '.'
-            if len(sentences) > 1:
-                response += ' ' + sentences[1] + '.'
-        else:
-            response = context[:300] + '...'
-
-        return response
-
-    def handle_conversational_interactions(self, query: str) -> Optional[str]:
-        """Handle comprehensive conversational interactions"""
-        query_lower = query.lower().strip()
-
-        # Use more specific patterns for greetings
-        greeting_patterns = [
-            r'^\s*(hello|hi|hey|hiya|howdy)\s*$',
-            r'^\s*(good morning|good afternoon|good evening|good day)\s*$',
-            r'^\s*(what\'s up|whats up|sup|yo)\s*$',
-            r'^\s*(greetings|salutations)\s*$',
-            r'^\s*(how are you|how are you doing|how\'s it going|hows it going)\s*$',
-            r'^\s*(good to meet you|nice to meet you|pleased to meet you)\s*$'
-        ]
-
-        for pattern in greeting_patterns:
-            if re.match(pattern, query_lower):
-                responses = [
-                    "👋 Hello! I'm BioGPT, your professional medical AI assistant specialized in pediatric medicine. I'm here to provide evidence-based medical information. What health concern can I help you with today?",
-                    "🏥 Hi there! I'm a medical AI assistant powered by BioGPT, trained on medical literature. I can help answer questions about children's health and medical conditions. How can I assist you?",
-                    "👋 Greetings! I'm your AI medical consultant, ready to help with pediatric health questions using the latest medical knowledge. What would you like to know about?"
-                ]
-                return np.random.choice(responses)
-
-        # ===== THANKS & APPRECIATION =====
-        thanks_patterns = [
-            ['thank you', 'thanks', 'thx', 'ty', 'thank you so much', 'thanks a lot', 'much appreciated', 'really appreciate it', 'i appreciate it', 'grateful', 'that was helpful', 'very helpful', 'awesome', 'perfect', 'great', 'excellent', 'wonderful', 'that helped', 'exactly what i needed', 'very informative', 'good information']
-        ]
-
-        for pattern_group in thanks_patterns:
-            if any(keyword in query_lower for keyword in pattern_group):
-                responses = [
-                    "🙏 You're very welcome! I'm glad I could provide helpful medical information. Remember, this is educational guidance - always consult your healthcare provider for personalized medical advice. Feel free to ask more questions!",
-                    "😊 Happy to help! Providing accurate medical information is what I'm here for. If you have any other pediatric health questions, don't hesitate to ask.",
-                    "🤗 You're most welcome! I'm pleased the medical information was useful. Please remember to consult with healthcare professionals for any medical decisions. What else can I help you with?"
-                ]
-                return np.random.choice(responses)
-
-        # ===== GOODBYES =====
-        goodbye_patterns = [
-            ['bye', 'goodbye', 'farewell', 'see you', 'later', 'see ya', 'catch you later', 'talk to you later', 'ttyl', 'have a good day', 'have a great day', 'take care', 'until next time', 'i need to go', 'that\'s all for now', 'no more questions']
-        ]
-
-        for pattern_group in goodbye_patterns:
-            if any(keyword in query_lower for keyword in pattern_group):
-                responses = [
-                    "👋 Goodbye! Take excellent care of yourself and your little ones. Remember, I'm here whenever you need reliable pediatric medical information. Stay healthy! 🏥",
-                    "🌟 Farewell! Wishing you and your family good health. Don't hesitate to return if you have more medical questions. Take care!",
-                    "👋 See you later! Hope the medical information was helpful. Remember to always consult healthcare professionals for medical decisions. Stay well!"
-                ]
-                return np.random.choice(responses)
-
-        # ===== ABOUT/HELP QUESTIONS =====
-        about_patterns = [
-            ['what are you', 'who are you', 'tell me about yourself', 'what do you do', 'what can you help with', 'what can you do', 'how can you help', 'what are your capabilities', 'help', 'help me', 'i need help', 'can you help', 'how do i use this', 'how does this work', 'what should i ask']
-        ]
-
-        for pattern_group in about_patterns:
-            if any(keyword in query_lower for keyword in pattern_group):
-                return """🤖 **About BioGPT Medical Assistant**
-
-I'm an AI medical assistant powered by BioGPT-Large, a specialized medical AI model trained on extensive medical literature. Here's what I can help you with:
-
-🩺 **Medical Specialties:**
-• Pediatric medicine and children's health
-• Symptom explanation and medical conditions
-• Treatment options and medical procedures
-• When to seek medical care
-• Prevention and wellness guidance
-
-🎯 **How to Use Me:**
-• Ask specific medical questions: "What causes fever in children?"
-• Describe symptoms: "My child has a persistent cough"
-• Seek guidance: "When should I call the doctor?"
-• Get information: "How do I treat dehydration?"
-
-⚠️ **Important Reminder:**
-I provide educational medical information based on medical literature, but I'm not a substitute for professional medical advice. Always consult qualified healthcare providers for:
-• Medical emergencies
-• Diagnosis and treatment decisions
-• Personalized medical advice
-• Medication guidance
-
-💡 **Tip:** Be specific in your questions for the most helpful responses!
-
-What pediatric health topic would you like to explore?"""
-
-        # ===== SMALL TALK & PERSONAL QUESTIONS =====
-        personal_patterns = [
-            ['how are you feeling', 'are you okay', 'how\'s your day', 'are you smart', 'are you intelligent', 'do you know everything', 'are you human', 'are you real', 'are you a robot', 'are you ai', 'you\'re smart', 'you\'re helpful', 'good job', 'well done', 'impressive']
-        ]
-
-        for pattern_group in personal_patterns:
-            if any(keyword in query_lower for keyword in pattern_group):
-                responses = [
-                    "🤖 I'm an AI medical assistant, so I don't have feelings, but I'm functioning well and ready to help with medical questions! My purpose is to provide reliable pediatric health information. What can I help you with?",
-                    "😊 Thank you for asking! As an AI, I'm always ready to assist with medical information. I'm designed to help with pediatric health questions using evidence-based medical knowledge. How can I help you today?",
-                    "🎯 I'm doing what I do best - providing medical information! I'm an AI trained on medical literature to help with pediatric health questions. What medical topic interests you?"
-                ]
-                return np.random.choice(responses)
-
-        # ===== CONFUSED/UNCLEAR INPUT =====
-        confusion_patterns = [
-            ['i don\'t know', 'not sure', 'confused', 'unclear', 'help me understand', 'what do you mean', 'i don\'t understand', 'can you explain', 'huh', 'i\'m lost', 'i\'m confused', 'this is confusing']
-        ]
-
-        for pattern_group in confusion_patterns:
-            if any(keyword in query_lower for keyword in pattern_group):
-                return """🤔 **I understand it can be confusing!** Let me help you get started.
-
-💡 **Try asking questions like:**
-
-🩺 **Symptoms:**
-• "What causes [symptom] in children?"
-• "My child has [symptom], what should I do?"
-
-💊 **Treatments:**
-• "How do I treat [condition] in children?"
-• "What are treatment options for [condition]?"
-
-🚨 **Urgency:**
-• "When should I call the doctor about [symptom]?"
-• "Is [symptom] serious in children?"
-
-🛡️ **Prevention:**
-• "How can I prevent [condition]?"
-• "What are the warning signs of [condition]?"
-
-**What specific aspect of your child's health would you like to understand better?**"""
-
-        # ===== APOLOGIES & POLITENESS =====
-        polite_patterns = [
-            ['sorry', 'excuse me', 'pardon me', 'my apologies', 'please help', 'could you please', 'would you mind', 'if you don\'t mind', 'sorry to bother you']
-        ]
-
-        for pattern_group in polite_patterns:
-            if any(keyword in query_lower for keyword in pattern_group):
-                return "😊 No need to apologize! I'm here to help with medical questions. Please feel free to ask anything about pediatric health - that's exactly what I'm designed for. What can I help you with?"
-
-        # ===== TESTING & VERIFICATION =====
-        test_patterns = [
-            ['test', 'testing', 'hello world', 'can you hear me', 'are you working', 'do you work', 'are you there', 'are you online', 'check', 'verify', 'ping']
-        ]
-
-        for pattern_group in test_patterns:
-            if any(keyword in query_lower for keyword in pattern_group):
-                return "✅ **System Check:** I'm working perfectly and ready to assist! BioGPT medical AI is online and functioning optimally. Ready to help with pediatric medical questions. What would you like to know?"
-
-        # Return None if no conversational pattern matches
-        return None
-
-
-    def chat(self, query: str) -> str:
-        """Main chat function with BioGPT and comprehensive conversational handling"""
-        if not query.strip():
-            return "Hello! I'm BioGPT, your professional medical AI assistant. How can I help you with pediatric medical questions today?"
-
-        # Handle comprehensive conversational interactions first
-        conversational_response = self.handle_conversational_interactions(query)
-        if conversational_response:
-            # Add to conversation history
-            self.conversation_history.append({
-                'query': query,
-                'response': conversational_response,
-                'timestamp': datetime.now().isoformat(),
-                'type': 'conversational'
-            })
-            return conversational_response
-
-        if not self.knowledge_chunks:
-            return "Please load medical data first to access the medical knowledge base."
-
-        if not self.model or not self.tokenizer:
-            return "Medical model not available. Please check the setup."
-
-        print(f"🔍 Processing medical query: {query}")
-
-        # Retrieve relevant medical context using FAISS or keyword search
-        start_time = time.time()
-        context = self.retrieve_medical_context(query)
-        retrieval_time = time.time() - start_time
-
-        if not context:
-            return "I don't have specific information about this topic in my medical database. Please consult with a healthcare professional for personalized medical advice."
-
-        print(f"   📚 Context retrieved ({retrieval_time:.2f}s)")
-
-        # Generate response with BioGPT
-        start_time = time.time()
-        main_context = '\n\n'.join(context)
-        response = self.generate_biogpt_response(main_context, query)
-        generation_time = time.time() - start_time
-
-        print(f"   🧠 Response generated ({generation_time:.2f}s)")
-
-        # Format final response
-        final_response = f"🩺 **Medical Information:** {response}\n\n⚠️ **Important:** This information is for educational purposes only. Always consult with qualified healthcare professionals for medical diagnosis, treatment, and personalized advice."
-
-        # Add to conversation history
-        self.conversation_history.append({
-            'query': query,
-            'response': final_response,
-            'timestamp': datetime.now().isoformat(),
-            'retrieval_time': retrieval_time,
-            'generation_time': generation_time,
-            'type': 'medical'
-        })
-
-        return final_response
-
-    def get_conversation_summary(self) -> Dict:
-        """Get conversation statistics"""
-        if not self.conversation_history:
-            return {"message": "No conversations yet"}
-
-        # Filter medical conversations for performance stats
-        medical_conversations = [h for h in self.conversation_history if h.get('type') == 'medical']
-
-        model_info = "BioGPT-Large" if self.model else "Model Not Loaded"
-
-        if not medical_conversations:
-            return {
-                "total_conversations": len(self.conversation_history),
-                "medical_conversations": 0,
-                "conversational_interactions": len(self.conversation_history),
-                "model_info": model_info,
-                "vector_search": "FAISS CPU" if self.faiss_ready else "Keyword Search",
-                "device": self.device
-            }
-
-        avg_retrieval_time = sum(h.get('retrieval_time', 0) for h in medical_conversations) / len(medical_conversations)
-        avg_generation_time = sum(h.get('generation_time', 0) for h in medical_conversations) / len(medical_conversations)
-
-        return {
-            "total_conversations": len(self.conversation_history),
-            "medical_conversations": len(medical_conversations),
-            "conversational_interactions": len(self.conversation_history) - len(medical_conversations),
-            "avg_retrieval_time": f"{avg_retrieval_time:.2f}s",
-            "avg_generation_time": f"{avg_generation_time:.2f}s",
-            "model_info": model_info,
-            "vector_search": "FAISS CPU" if self.faiss_ready else "Keyword Search",
-            "device": self.device,
-            "quantization": "8-bit" if self.use_8bit else "16-bit/32-bit"
-        }
-
-# Create and Test BioGPT Chatbot
-
-def create_biogpt_chatbot():
-    """Create and initialize the BioGPT chatbot"""
-    print("🚀 Creating Professional BioGPT Medical Chatbot")
-    print("=" * 60)
-
-    # Create chatbot
-    chatbot = ColabBioGPTChatbot(use_gpu=True, use_8bit=True)
-
-    return chatbot
-
-def test_biogpt_chatbot(chatbot, test_file='Pediatric_cleaned.txt'):
-    """Test the BioGPT chatbot"""
-    print("\n📚 Loading medical data...")
-    success = chatbot.load_medical_data(test_file)
-
-    if not success:
-        print("❌ Failed to load medical data. Please check the file.")
-        return None
-
-    print("\n🧪 Testing BioGPT Medical Chatbot:")
-    print("=" * 50)
-
-    # Test queries
-    test_queries = [
-        "What causes fever in children?",
-        "How should I treat my child's cough?",
-        "When should I be concerned about my baby's breathing?",
-        "What are the signs of dehydration in infants?"
-    ]
-
-    for i, query in enumerate(test_queries, 1):
-        print(f"\n{i}️⃣ Testing: {query}")
-        print("-" * 40)
-
-        response = chatbot.chat(query)
-        print(f"🤖 BioGPT Response:\n{response}")
-        print("=" * 50)
-
-    # Show conversation summary
-    summary = chatbot.get_conversation_summary()
-    print("\n📊 Performance Summary:")
-    for key, value in summary.items():
-        print(f"   {key}: {value}")
-
-    return chatbot
-
-# Interactive Chat Interface
-
-def interactive_biogpt_chat(chatbot):
-    """Interactive chat with BioGPT"""
-    print("\n💬 Interactive BioGPT Medical Chat")
-    print("=" * 50)
-    print("You're now chatting with BioGPT, a professional medical AI!")
-    print("Type 'quit' to exit, 'summary' to see stats")
-    print("-" * 50)
-
-    while True:
-        user_input = input("\n👤 You: ").strip()
-
-        if user_input.lower() in ['quit', 'exit', 'bye']:
-            print("\n👋 Thank you for using BioGPT Medical Assistant!")
-            # Show final summary
-            summary = chatbot.get_conversation_summary()
-            print("\n📊 Final Session Summary:")
-            for key, value in summary.items():
-                print(f"   {key}: {value}")
-            break
-
-        elif user_input.lower() == 'summary':
-            summary = chatbot.get_conversation_summary()
-            print("\n📊 Current Session Summary:")
-            for key, value in summary.items():
-                print(f"   {key}: {value}")
-            continue
-
-        elif not user_input:
-            continue
-
-        print(f"\n🤖 BioGPT: ", end="")
-        response = chatbot.chat(user_input)
-        print(response)
-
-# Main Execution - Initialization
-
-import torch
-
-# Create the BioGPT chatbot
-chatbot = create_biogpt_chatbot()
-
-print("\n" + "="*60)
-print("🎯 NEXT STEPS:")
-print("1. Upload your medical data file by running the next cell.")
-print("2. Test the chatbot by running the cell after the upload.")
-print("3. Start interactive chat by running the final cell.")
-print("="*60)
-