app.py

import os
import sys
import json
import logging
import warnings
from pathlib import Path
from typing import List, Dict, Any, Optional, Tuple
import hashlib
import pickle
from datetime import datetime
import time
import asyncio
from concurrent.futures import ThreadPoolExecutor

# Suppress warnings for cleaner output
warnings.filterwarnings("ignore")

# Core dependencies
import gradio as gr
import numpy as np
import pandas as pd
from sentence_transformers import SentenceTransformer
import faiss
import torch
from transformers import (
    AutoTokenizer, 
    AutoModelForCausalLM, 
    BitsAndBytesConfig,
    pipeline
)

# Document processing
from llama_index.core import Document, VectorStoreIndex, Settings
from llama_index.core.node_parser import SentenceSplitter
from llama_index.vector_stores.faiss import FaissVectorStore
from llama_index.embeddings.huggingface import HuggingFaceEmbedding
from llama_index.core import StorageContext

# PDF processing
import PyPDF2
from io import BytesIO

# Medical knowledge validation
import re

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
)
logger = logging.getLogger(__name__)

class MedicalFactChecker:
    """Enhanced medical fact checker with faster validation"""
    
    def __init__(self):
        self.medical_facts = self._load_medical_facts()
        self.contraindications = self._load_contraindications()
        self.dosage_patterns = self._compile_dosage_patterns()
        self.definitive_patterns = [
            re.compile(r, re.IGNORECASE) for r in [
                r'always\s+(?:use|take|apply)',
                r'never\s+(?:use|take|apply)',
                r'will\s+(?:cure|heal|fix)',
                r'guaranteed\s+to',
                r'completely\s+(?:safe|effective)'
            ]
        ]
        
    def _load_medical_facts(self) -> Dict[str, Any]:
        """Pre-loaded medical facts for Gaza context"""
        return {
            "burn_treatment": {
                "cool_water": "Use clean, cool (not ice-cold) water for 10-20 minutes",
                "no_ice": "Never apply ice directly to burns",
                "clean_cloth": "Cover with clean, dry cloth if available"
            },
            "wound_care": {
                "pressure": "Apply direct pressure to control bleeding",
                "elevation": "Elevate injured limb if possible",
                "clean_hands": "Clean hands before treating wounds when possible"
            },
            "infection_signs": {
                "redness": "Increasing redness around wound",
                "warmth": "Increased warmth at wound site",
                "pus": "Yellow or green discharge",
                "fever": "Fever may indicate systemic infection"
            }
        }
    
    def _load_contraindications(self) -> Dict[str, List[str]]:
        """Pre-loaded contraindications for common treatments"""
        return {
            "aspirin": ["children under 16", "bleeding disorders", "stomach ulcers"],
            "ibuprofen": ["kidney disease", "heart failure", "stomach bleeding"],
            "hydrogen_peroxide": ["deep wounds", "closed wounds", "eyes"],
            "tourniquets": ["non-life-threatening bleeding", "without proper training"]
        }
    
    def _compile_dosage_patterns(self) -> List[re.Pattern]:
        """Pre-compiled dosage patterns"""
        patterns = [
            r'\d+\s*mg\b',  # milligrams
            r'\d+\s*g\b',   # grams
            r'\d+\s*ml\b',  # milliliters
            r'\d+\s*tablets?\b',  # tablets
            r'\d+\s*times?\s+(?:per\s+)?day\b',  # frequency
            r'every\s+\d+\s+hours?\b'  # intervals
        ]
        return [re.compile(pattern, re.IGNORECASE) for pattern in patterns]
    
    def check_medical_accuracy(self, response: str, context: str) -> Dict[str, Any]:
        """Enhanced medical accuracy check with Gaza-specific considerations"""
        issues = []
        warnings = []
        accuracy_score = 0.0
        
        # Check for contraindications (faster keyword matching)
        response_lower = response.lower()
        for medication, contra_list in self.contraindications.items():
            if medication in response_lower:
                for contra in contra_list:
                    if any(word in response_lower for word in contra.split()):
                        issues.append(f"Potential contraindication: {medication} with {contra}")
                        accuracy_score -= 0.3
                        break
        
        # Context alignment using Jaccard similarity
        if context:
            resp_words = set(response_lower.split())
            ctx_words = set(context.lower().split())
            context_similarity = len(resp_words & ctx_words) / len(resp_words | ctx_words) if ctx_words else 0.0
            if context_similarity < 0.5:  # Lowered threshold for Gaza context
                warnings.append(f"Low context similarity: {context_similarity:.2f}")
                accuracy_score -= 0.1
        else:
            context_similarity = 0.0
        
        # Gaza-specific resource checks
        gaza_resources = ["clean water", "sterile", "hospital", "ambulance", "electricity"]
        if any(resource in response_lower for resource in gaza_resources):
            warnings.append("Consider resource limitations in Gaza context")
            accuracy_score -= 0.05
        
        # Unsupported claims check
        for pattern in self.definitive_patterns:
            if pattern.search(response):
                issues.append(f"Unsupported definitive claim detected")
                accuracy_score -= 0.4
                break
        
        # Dosage validation
        for pattern in self.dosage_patterns:
            if pattern.search(response):
                warnings.append("Dosage detected - verify with professional")
                accuracy_score -= 0.1
                break
        
        confidence_score = max(0.0, min(1.0, 0.8 + accuracy_score))
        
        return {
            "confidence_score": confidence_score,
            "issues": issues,
            "warnings": warnings,
            "context_similarity": context_similarity,
            "is_safe": len(issues) == 0 and confidence_score > 0.5
        }

class EnhancedGazaKnowledgeBase:
    """Enhanced knowledge base with better embeddings and indexing"""
    
    def __init__(self, data_dir: str = "./data"):
        self.data_dir = Path(data_dir)
        self.embedding_model = None
        self.vector_store = None
        self.index = None
        self.chunk_metadata = []
        self.index_path = self.data_dir / "enhanced_vector_store"
        
        # Enhanced medical priorities for Gaza context
        self.medical_priorities = {
            "trauma": ["gunshot", "blast", "burns?", "fracture", "shrapnel", "explosion"],
            "infectious": ["cholera", "dysentery", "infection", "sepsis", "wound infection"],
            "chronic": ["diabetes", "hypertension", "malnutrition", "kidney", "heart"],
            "emergency": ["cardiac", "bleeding", "airway", "unconscious", "shock"],
            "gaza_specific": ["siege", "blockade", "limited supplies", "no electricity", "water shortage"]
        }
        
    def initialize(self):
        """Enhanced initialization with better embedding model"""
        if not self.index_path.exists():
            self.index_path.mkdir(parents=True)
            
        # Use a more powerful medical embedding model
        device = 'cuda' if torch.cuda.is_available() else 'cpu'
        
        # Try to use a medical-specific embedding model, fallback to general model
        try:
            # First try a medical-specific model (if available)
            self.embedding_model = HuggingFaceEmbedding(
                model_name="sentence-transformers/all-mpnet-base-v2",  # Higher dimension (768)
                device=device,
                embed_batch_size=4
            )
            logger.info("Using all-mpnet-base-v2 (768-dim) embedding model")
        except Exception as e:
            logger.warning(f"Failed to load preferred model, using fallback: {e}")
            self.embedding_model = HuggingFaceEmbedding(
                model_name="sentence-transformers/all-MiniLM-L6-v2",
                device=device,
                embed_batch_size=4
            )
            logger.info("Using all-MiniLM-L6-v2 (384-dim) embedding model")
        
        # Configure global settings
        Settings.embed_model = self.embedding_model
        Settings.chunk_size = 512  # Increased chunk size for better context
        Settings.chunk_overlap = 50  # Increased overlap
        
        # Check for existing index
        if (self.index_path / "index.faiss").exists() and (self.index_path / "docstore.json").exists():
            self._load_vector_store()
        else:
            self._create_vector_store()



    def _batch_embed_with_retry(self, texts, batch_size=16, max_retries=3, delay=2):
        """
        Embed texts in batches with retry fallback and logging
        """
        embeddings = []
        for i in range(0, len(texts), batch_size):
            batch = texts[i:i+batch_size]
        for attempt in range(max_retries):
            try:
                batch_embeddings = self.embedding_model.get_text_embedding_batch(batch)
                embeddings.extend(batch_embeddings)
                break  # success
            except Exception as e:
                if attempt < max_retries - 1:
                    logger.warning(f"Batch {i}-{i+len(batch)} failed (attempt {attempt+1}): {e}. Retrying...")
                    time.sleep(delay * (attempt + 1))
                else:
                    logger.error(f"❌ Final failure embedding batch {i}-{i+len(batch)}: {e}")
                    for text in batch:
                        try:
                            embeddings.append(self.embedding_model.get_text_embedding(text))
                        except Exception as sub_e:
                            logger.error(f"Failed to embed single text: {sub_e} — {text[:60]}...")
        return embeddings



    
    def _load_vector_store(self):
        """Load existing vector store with error handling"""
        try:
            # Load the FAISS index directly
            faiss_index = faiss.read_index(str(self.index_path / "index.faiss"))
            vector_store = FaissVectorStore(faiss_index=faiss_index)
            
            # Create storage context
            storage_context = StorageContext.from_defaults(
                vector_store=vector_store,
                persist_dir=str(self.index_path)
            )
            
            # Load the index
            self.index = VectorStoreIndex.load(
                storage_context=storage_context
            )
            
            # Load metadata
            metadata_path = self.index_path / "metadata.pkl"
            if metadata_path.exists():
                with open(metadata_path, 'rb') as f:
                    self.chunk_metadata = pickle.load(f)
            
            logger.info(f"Loaded existing vector store with {len(self.chunk_metadata)} chunks")
        except Exception as e:
            logger.error(f"Error loading vector store: {e}")
            # Fallback to creating new store if loading fails
            self._create_vector_store()
    
    def _create_vector_store(self):
        """Create enhanced vector store with IVF indexing"""
        documents = self._load_documents()
        
        if not documents:
            logger.warning("No documents found. Creating empty index")
            self.chunk_metadata = []
            return
        
        # Determine embedding dimension
        try:
            test_embedding = self.embedding_model.get_text_embedding("test")
            dimension = len(test_embedding)
            logger.info(f"Embedding dimension: {dimension}")
        except Exception as e:
            logger.error(f"Failed to determine embedding dimension: {e}")
            dimension = 768  # Default for all-mpnet-base-v2
        
        # Create enhanced FAISS index with IVF for better performance
        try:
            # For small datasets, use flat index; for larger ones, use IVF
            if len(documents) < 1000:
                faiss_index = faiss.IndexFlatL2(dimension)
                logger.info("Using IndexFlatL2 for small dataset")
            else:
                # Use IVF with reasonable number of clusters
                nlist = min(100, len(documents) // 10)  # Adaptive cluster count
                quantizer = faiss.IndexFlatL2(dimension)
                faiss_index = faiss.IndexIVFFlat(quantizer, dimension, nlist)
                logger.info(f"Using IndexIVFFlat with {nlist} clusters")
        except Exception as e:
            logger.error(f"Failed to create enhanced index, using flat: {e}")
            faiss_index = faiss.IndexFlatL2(dimension)
        
        vector_store = FaissVectorStore(faiss_index=faiss_index)
        
        # Create storage context
        storage_context = StorageContext.from_defaults(
            vector_store=vector_store
        )
        
        # Configure node parser with enhanced settings
        parser = SentenceSplitter(
            chunk_size=Settings.chunk_size,
            chunk_overlap=Settings.chunk_overlap,
            include_prev_next_rel=True  # Include relationships for better context
        )
        
        # Create index using global settings
        self.index = VectorStoreIndex.from_documents(
            documents,
            storage_context=storage_context,
            transformations=[parser],
            show_progress=True
        )
        
        # Train IVF index if needed
        if hasattr(faiss_index, 'train') and not faiss_index.is_trained:
            logger.info("Training IVF index...")
            # Get some embeddings for training
            sample_texts = [doc.text[:500] for doc in documents[:100]]  # Sample for training
            sample_embeddings = np.array(self._batch_embed_with_retry(sample_texts, batch_size=16)).astype('float32')
            faiss_index.train(sample_embeddings)
            logger.info("IVF index training completed")
        
        # Save metadata
        self.chunk_metadata = [
            {"text": node.text, "source": node.metadata.get("source", "unknown")}
            for node in self.index.docstore.docs.values()
        ]
        
        # Persist the index
        self.index.storage_context.persist(persist_dir=str(self.index_path))
        
        # Save metadata separately
        with open(self.index_path / "metadata.pkl", 'wb') as f:
            pickle.dump(self.chunk_metadata, f)
        
        logger.info(f"Created enhanced vector store with {len(self.chunk_metadata)} chunks")
        
    def _load_documents(self) -> List[Document]:
        """Enhanced document loading with better caching"""
        documents = []
        doc_cache = self.index_path / "document_cache.pkl"
        
        # Try loading from cache
        if doc_cache.exists():
            try:
                with open(doc_cache, 'rb') as f:
                    cached_data = pickle.load(f)
                    if isinstance(cached_data, dict) and 'documents' in cached_data:
                        cached_docs = cached_data['documents']
                        if isinstance(cached_docs, list) and all(isinstance(d, Document) for d in cached_docs):
                            logger.info(f"Loaded {len(cached_docs)} documents from cache")
                            return cached_docs
                    logger.warning("Document cache format invalid")
            except Exception as e:
                logger.warning(f"Document cache corrupted: {e}")
        
        # Process files with enhanced error handling
        processed_files = []
        for pdf_file in self.data_dir.glob("*.pdf"):
            try:
                doc_text = self._extract_pdf_text(pdf_file)
                if doc_text and len(doc_text.strip()) > 100:  # Minimum content check
                    documents.append(Document(
                        text=doc_text,
                        metadata={
                            "source": str(pdf_file.name), 
                            "type": "pdf",
                            "file_size": pdf_file.stat().st_size,
                            "processed_date": datetime.now().isoformat()
                        }
                    ))
                    processed_files.append(str(pdf_file.name))
                    logger.info(f"Processed: {pdf_file.name} ({len(doc_text)} chars)")
            except Exception as e:
                logger.error(f"Error loading {pdf_file}: {e}")
        
        # Process text files as well
        for txt_file in self.data_dir.glob("*.txt"):
            try:
                with open(txt_file, 'r', encoding='utf-8') as f:
                    doc_text = f.read()
                    if doc_text and len(doc_text.strip()) > 100:
                        documents.append(Document(
                            text=doc_text,
                            metadata={
                                "source": str(txt_file.name), 
                                "type": "txt",
                                "file_size": txt_file.stat().st_size,
                                "processed_date": datetime.now().isoformat()
                            }
                        ))
                        processed_files.append(str(txt_file.name))
                        logger.info(f"Processed: {txt_file.name} ({len(doc_text)} chars)")
            except Exception as e:
                logger.error(f"Error loading {txt_file}: {e}")
        
        # Save to cache if we found documents
        if documents:
            cache_data = {
                'documents': documents,
                'processed_files': processed_files,
                'cache_date': datetime.now().isoformat()
            }
            with open(doc_cache, 'wb') as f:
                pickle.dump(cache_data, f)
            logger.info(f"Cached {len(documents)} documents")
            
        return documents
    
    def _extract_pdf_text(self, pdf_path: Path) -> str:
        """Enhanced PDF text extraction with better error handling"""
        try:
            with open(pdf_path, 'rb') as file:
                pdf_reader = PyPDF2.PdfReader(file)
                text = []
                
                for page_num, page in enumerate(pdf_reader.pages):
                    try:
                        page_text = page.extract_text()
                        if page_text and page_text.strip():
                            # Clean up the text
                            page_text = re.sub(r'\s+', ' ', page_text)  # Normalize whitespace
                            text.append(page_text)
                    except Exception as e:
                        logger.warning(f"Error extracting page {page_num} from {pdf_path}: {e}")
                        continue
                
                full_text = "\n".join(text) if text else ""
                
                # Additional validation
                if len(full_text.strip()) < 100:
                    logger.warning(f"Extracted text too short from {pdf_path}")
                    return ""
                    
                return full_text
                
        except Exception as e:
            logger.error(f"Error extracting text from {pdf_path}: {e}")
            return ""
    
    def search(self, query: str, k: int = 5) -> List[Dict[str, Any]]:
        """Enhanced search with better error handling and result processing"""
        if not self.index:
            logger.warning("Index not available for search")
            return []
            
        try:
            retriever = self.index.as_retriever(similarity_top_k=k)
            results = retriever.retrieve(query)
            
            # FIX: Handle the tuple object error by properly extracting node and score
            processed_results = []
            for result in results:
                try:
                    # Handle both tuple and direct node results
                    if isinstance(result, tuple):
                        node, score = result
                    else:
                        node = result
                        score = getattr(result, 'score', 0.0)
                    
                    # Extract text safely
                    text = getattr(node, 'text', str(node))
                    source = node.metadata.get("source", "unknown") if hasattr(node, 'metadata') else "unknown"
                    
                    processed_results.append({
                        "text": text,
                        "source": source,
                        "score": float(score) if score is not None else 0.0,
                        "medical_priority": self._assess_priority(text)
                    })
                except Exception as e:
                    logger.error(f"Error processing search result: {e}")
                    continue
            
            # Sort by score (higher is better)
            processed_results.sort(key=lambda x: x['score'], reverse=True)
            
            logger.info(f"Search returned {len(processed_results)} results for query: {query[:50]}...")
            return processed_results
            
        except Exception as e:
            logger.error(f"Error during search: {e}")
            return []
    
    def _assess_priority(self, text: str) -> str:
        """Enhanced medical priority assessment"""
        text_lower = text.lower()
        
        # Check priorities in order of importance
        priority_order = ["emergency", "trauma", "gaza_specific", "infectious", "chronic"]
        
        for priority in priority_order:
            keywords = self.medical_priorities.get(priority, [])
            if any(re.search(keyword, text_lower) for keyword in keywords):
                return priority
        
        return "general"

class EnhancedGazaRAGSystem:
    """Enhanced RAG system with better performance and error handling"""
    
    def __init__(self):
        self.knowledge_base = EnhancedGazaKnowledgeBase()
        self.fact_checker = MedicalFactChecker()
        self.llm = None
        self.tokenizer = None
        self.system_prompt = self._create_system_prompt()
        self.generation_pipeline = None
        self.response_cache = {}  # Simple response caching
        self.executor = ThreadPoolExecutor(max_workers=2)  # For async processing
        
    def initialize(self):
        """Enhanced initialization with better error handling"""
        logger.info("Initializing Enhanced Gaza RAG System...")
        
        try:
            self.knowledge_base.initialize()
            logger.info("Knowledge base initialized successfully")
        except Exception as e:
            logger.error(f"Failed to initialize knowledge base: {e}")
            raise
        
        # Lazy LLM loading - will load on first request
        logger.info("RAG system ready (LLM will load on first request)")
    
    def _initialize_llm(self):
        """Enhanced LLM initialization with better error handling"""
        if self.llm is not None:
            return
            
        model_name = "microsoft/Phi-3-mini-4k-instruct"
        try:
            logger.info(f"Loading LLM: {model_name}")
            
            # Enhanced quantization configuration
            quantization_config = BitsAndBytesConfig(
                load_in_4bit=True,
                bnb_4bit_compute_dtype=torch.float16,
                bnb_4bit_use_double_quant=True,
                bnb_4bit_quant_type="nf4",
                bnb_4bit_quant_storage=torch.uint8
            )
            
            self.tokenizer = AutoTokenizer.from_pretrained(
                model_name,
                trust_remote_code=True,
                padding_side="left"  # Better for generation
            )
            
            # Add pad token if missing
            if self.tokenizer.pad_token is None:
                self.tokenizer.pad_token = self.tokenizer.eos_token
            
            self.llm = AutoModelForCausalLM.from_pretrained(
                model_name,
                quantization_config=quantization_config,
                device_map="auto",
                trust_remote_code=True,
                torch_dtype=torch.float16,
                low_cpu_mem_usage=True
            )
            
            # Create enhanced pipeline
            self.generation_pipeline = pipeline(
                "text-generation",
                model=self.llm,
                tokenizer=self.tokenizer,
                device_map="auto",
                torch_dtype=torch.float16,
                return_full_text=False  # Only return generated text
            )
            
            logger.info("LLM loaded successfully")
            
        except Exception as e:
            logger.error(f"Error loading primary model: {e}")
            self._initialize_fallback_llm()
    
    def _initialize_fallback_llm(self):
        """Enhanced fallback model with better error handling"""
        try:
            logger.info("Loading fallback model...")
            
            fallback_model = "microsoft/DialoGPT-small"
            self.tokenizer = AutoTokenizer.from_pretrained(fallback_model)
            self.llm = AutoModelForCausalLM.from_pretrained(
                fallback_model,
                torch_dtype=torch.float32,
                low_cpu_mem_usage=True
            )
            
            if self.tokenizer.pad_token is None:
                self.tokenizer.pad_token = self.tokenizer.eos_token
            
            self.generation_pipeline = pipeline(
                "text-generation",
                model=self.llm,
                tokenizer=self.tokenizer,
                return_full_text=False
            )
            
            logger.info("Fallback model loaded successfully")
            
        except Exception as e:
            logger.error(f"Fallback model failed: {e}")
            self.llm = None
            self.generation_pipeline = None
    
    def _create_system_prompt(self) -> str:
        """Enhanced system prompt for Gaza context"""
        return """You are a medical AI assistant specifically designed for Gaza healthcare workers operating under siege conditions. 

CRITICAL GUIDELINES:
- Provide practical first aid guidance considering limited resources (water, electricity, medical supplies)
- Always prioritize patient safety and recommend professional medical help when available
- Consider Gaza's specific challenges: blockade, limited hospitals, frequent power outages
- Suggest alternative treatments when standard medical supplies are unavailable
- Never provide definitive diagnoses - only supportive care guidance
- Be culturally sensitive and aware of the humanitarian crisis context

RESOURCE CONSTRAINTS TO CONSIDER:
- Limited clean water availability
- Frequent electricity outages
- Restricted medical supply access
- Overwhelmed healthcare facilities
- Limited transportation for medical emergencies

Provide clear, actionable advice while emphasizing the need for professional medical care when possible."""
    
    async def generate_response_async(self, query: str, progress_callback=None) -> Dict[str, Any]:
        """Async response generation with progress tracking"""
        start_time = time.time()
        
        if progress_callback:
            progress_callback(0.1, "Checking cache...")
        
        # Check cache first
        query_hash = hashlib.md5(query.encode()).hexdigest()
        if query_hash in self.response_cache:
            cached_response = self.response_cache[query_hash]
            cached_response["cached"] = True
            cached_response["response_time"] = 0.1
            if progress_callback:
                progress_callback(1.0, "Retrieved from cache!")
            return cached_response
        
        try:
            if progress_callback:
                progress_callback(0.2, "Initializing LLM...")
            
            # Initialize LLM only when needed
            if self.llm is None:
                await asyncio.get_event_loop().run_in_executor(
                    self.executor, self._initialize_llm
                )
            
            if progress_callback:
                progress_callback(0.4, "Searching knowledge base...")
                
            # Enhanced knowledge retrieval
            search_results = await asyncio.get_event_loop().run_in_executor(
                self.executor, self.knowledge_base.search, query, 3
            )
            
            if progress_callback:
                progress_callback(0.6, "Preparing context...")
            
            context = self._prepare_context(search_results)
            
            if progress_callback:
                progress_callback(0.8, "Generating response...")
            
            # Generate response
            response = await asyncio.get_event_loop().run_in_executor(
                self.executor, self._generate_response, query, context
            )
            
            if progress_callback:
                progress_callback(0.9, "Validating safety...")
            
            # Enhanced safety check
            safety_check = self.fact_checker.check_medical_accuracy(response, context)
            
            # Prepare final response
            final_response = self._prepare_final_response(
                response, 
                search_results, 
                safety_check,
                time.time() - start_time
            )
            
            # Cache the response (limit cache size)
            if len(self.response_cache) < 100:
                self.response_cache[query_hash] = final_response
            
            if progress_callback:
                progress_callback(1.0, "Complete!")
            
            return final_response
            
        except Exception as e:
            logger.error(f"Error generating response: {e}")
            if progress_callback:
                progress_callback(1.0, f"Error: {str(e)}")
            return self._create_error_response(str(e))
    
    
    
    
    def _generate_response(self, query: str, context: str) -> str:
        """Enhanced response generation using model.generate() to avoid DynamicCache errors"""
        if self.llm is None or self.tokenizer is None:
            return self._generate_fallback_response(query, context)

        # Build prompt with Gaza-specific context
        prompt = f"""{self.system_prompt}
        MEDICAL KNOWLEDGE CONTEXT:
        {context}
        PATIENT QUESTION: {query}
        RESPONSE (provide practical, Gaza-appropriate medical guidance):"""

        try:
            # Tokenize and move to correct device
            inputs = self.tokenizer(prompt, return_tensors="pt").to(self.llm.device)

            # Generate the response
            outputs = self.llm.generate(
                **inputs,
                max_new_tokens=800,
                temperature=0.5,
                pad_token_id=self.tokenizer.eos_token_id,
                do_sample=True,
                repetition_penalty=1.15,
            )

            # Decode and clean up
            response_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
            lines = response_text.split('\n')
            unique_lines = []
            for line in lines:
                line = line.strip()
                if line and line not in unique_lines:
                    unique_lines.append(line)
            return '\n'.join(unique_lines)

        except Exception as e:
            logger.error(f"Error in LLM generate(): {e}")
            return self._generate_fallback_response(query, context)


        # Decode and clean up
        response_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
        lines = response_text.split('\n')
        unique_lines = []
        for line in lines:
            line = line.strip()
            if line and line not in unique_lines:
                unique_lines.append(line)
        return '\n'.join(unique_lines)


    
    
    def _prepare_context(self, search_results: List[Dict[str, Any]]) -> str:
        """Enhanced context preparation with better formatting"""
        if not search_results:
            return "No specific medical guidance found in knowledge base. Provide general first aid principles."
        
        context_parts = []
        for i, result in enumerate(search_results, 1):
            source = result.get('source', 'unknown')
            text = result.get('text', '')
            priority = result.get('medical_priority', 'general')
            
            # Truncate long text but preserve important information
            if len(text) > 400:
                text = text[:400] + "..."
            
            context_parts.append(f"[Source {i}: {source} - Priority: {priority}]\n{text}")
        
        return "\n\n".join(context_parts)
    
    def _generate_response(self, query: str, context: str) -> str:
        """Enhanced response generation with better prompting"""
        if not self.generation_pipeline:
            return self._generate_fallback_response(query, context)
        
        # Enhanced prompt structure
        prompt = f"""{self.system_prompt}

MEDICAL KNOWLEDGE CONTEXT:
{context}

PATIENT QUESTION: {query}

RESPONSE (provide practical, Gaza-appropriate medical guidance):"""
        
        try:
            # Enhanced generation parameters
            response = self.generation_pipeline(
                prompt,
                max_new_tokens=300,  # Increased for more detailed responses
                temperature=0.2,     # Lower for more consistent medical advice
                do_sample=True,
                pad_token_id=self.tokenizer.eos_token_id,
                repetition_penalty=1.15,
                truncation=True,
                num_return_sequences=1
            )
            
            if response and len(response) > 0:
                generated_text = response[0]['generated_text']
                # Clean up the response
                generated_text = generated_text.strip()
                
                # Remove any repetitive patterns
                lines = generated_text.split('\n')
                unique_lines = []
                for line in lines:
                    if line.strip() and line.strip() not in unique_lines:
                        unique_lines.append(line.strip())
                
                return '\n'.join(unique_lines)
            else:
                return self._generate_fallback_response(query, context)
                
        except Exception as e:
            logger.error(f"Error in LLM generation: {e}")
            return self._generate_fallback_response(query, context)
    
    def _generate_fallback_response(self, query: str, context: str) -> str:
        """Enhanced fallback response with Gaza-specific guidance"""
        gaza_guidance = {
            "burn": "For burns: Use clean, cool water if available. If water is scarce, use clean cloth. Avoid ice. Seek medical help urgently.",
            "bleeding": "For bleeding: Apply direct pressure with clean cloth. Elevate if possible. If severe, seek immediate medical attention.",
            "wound": "For wounds: Clean hands if possible. Apply pressure to stop bleeding. Cover with clean material. Watch for infection signs.",
            "infection": "Signs of infection: Redness, warmth, swelling, pus, fever. Seek medical care immediately if available.",
            "pain": "For pain management: Rest, elevation, cold/warm compress as appropriate. Avoid aspirin in children."
        }
        
        query_lower = query.lower()
        for condition, guidance in gaza_guidance.items():
            if condition in query_lower:
                return f"{guidance}\n\nContext from medical sources:\n{context[:200]}..."
        
        return f"Medical guidance for: {query}\n\nGeneral advice: Prioritize safety, seek professional help when available, consider resource limitations in Gaza.\n\nRelevant information:\n{context[:300]}..."
    
    def _prepare_final_response(
        self, 
        response: str, 
        search_results: List[Dict[str, Any]], 
        safety_check: Dict[str, Any],
        response_time: float
    ) -> Dict[str, Any]:
        """Enhanced final response preparation with more metadata"""
        
        # Add safety warnings if needed
        if not safety_check["is_safe"]:
            response = f"⚠️ MEDICAL CAUTION: {response}\n\n🚨 Please verify this guidance with a medical professional when possible."
        
        # Add Gaza-specific disclaimer
        response += "\n\n📍 Gaza Context: This guidance considers resource limitations. Adapt based on available supplies and seek professional medical care when accessible."
        
        # Extract unique sources
        sources = list(set(res.get("source", "unknown") for res in search_results)) if search_results else []
        
        # Calculate confidence based on multiple factors
        base_confidence = safety_check.get("confidence_score", 0.5)
        context_bonus = 0.1 if search_results else 0.0
        safety_penalty = 0.2 if not safety_check.get("is_safe", True) else 0.0
        
        final_confidence = max(0.0, min(1.0, base_confidence + context_bonus - safety_penalty))
        
        return {
            "response": response,
            "confidence": final_confidence,
            "sources": sources,
            "search_results_count": len(search_results),
            "safety_issues": safety_check.get("issues", []),
            "safety_warnings": safety_check.get("warnings", []),
            "response_time": round(response_time, 2),
            "timestamp": datetime.now().isoformat()[:19],
            "cached": False
        }
    
    def _create_error_response(self, error_msg: str) -> Dict[str, Any]:
        """Enhanced error response with helpful information"""
        return {
            "response": f"⚠️ System Error: Unable to process your medical query at this time.\n\nError: {error_msg}\n\n🚨 For immediate medical emergencies, seek professional help directly.\n\n📞 Gaza Emergency Numbers:\n- Palestinian Red Crescent: 101\n- Civil Defense: 102",
            "confidence": 0.0,
            "sources": [],
            "search_results_count": 0,
            "safety_issues": ["System error occurred"],
            "safety_warnings": ["Unable to validate medical accuracy"],
            "response_time": 0.0,
            "timestamp": datetime.now().isoformat()[:19],
            "cached": False,
            "error": True
        }

# Global system instance
enhanced_rag_system = None

def initialize_enhanced_system():
    """Initialize enhanced system with better error handling"""
    global enhanced_rag_system
    if enhanced_rag_system is None:
        try:
            enhanced_rag_system = EnhancedGazaRAGSystem()
            enhanced_rag_system.initialize()
            logger.info("Enhanced Gaza RAG System initialized successfully")
        except Exception as e:
            logger.error(f"Failed to initialize enhanced system: {e}")
            raise
    return enhanced_rag_system

def process_medical_query_with_progress(query: str, progress=gr.Progress()) -> Tuple[str, str, str]:
    """Enhanced query processing with detailed progress tracking and status updates"""
    if not query.strip():
        return "Please enter a medical question.", "", "⚠️ No query provided"
    
    try:
        # Initialize system with progress
        progress(0.05, desc="🔧 Initializing system...")
        system = initialize_enhanced_system()
        
        # Create async event loop for progress tracking
        loop = asyncio.new_event_loop()
        asyncio.set_event_loop(loop)
        
        def progress_callback(value, desc):
            progress(value, desc=desc)
        
        try:
            # Run async generation with progress
            result = loop.run_until_complete(
                system.generate_response_async(query, progress_callback)
            )
        finally:
            loop.close()
        
        # Prepare response with enhanced metadata
        response = result["response"]
        
        # Prepare detailed metadata
        metadata_parts = [
            f"🎯 Confidence: {result['confidence']:.1%}",
            f"⏱️ Response: {result['response_time']}s",
            f"📚 Sources: {result['search_results_count']} found"
        ]
        
        if result.get('cached'):
            metadata_parts.append("💾 Cached")
        
        if result.get('sources'):
            metadata_parts.append(f"📖 Refs: {', '.join(result['sources'][:2])}")
        
        metadata = " | ".join(metadata_parts)
        
        # Prepare status with warnings/issues
        status_parts = []
        if result.get('safety_warnings'):
            status_parts.append(f"⚠️ {len(result['safety_warnings'])} warnings")
        if result.get('safety_issues'):
            status_parts.append(f"🚨 {len(result['safety_issues'])} issues")
        if not status_parts:
            status_parts.append("✅ Safe response")
        
        status = " | ".join(status_parts)
        
        return response, metadata, status
        
    except Exception as e:
        logger.error(f"Error processing query: {e}")
        error_response = f"⚠️ Error processing your query: {str(e)}\n\n🚨 For medical emergencies, seek immediate professional help."
        error_metadata = f"❌ Error at {datetime.now().strftime('%H:%M:%S')}"
        error_status = "🚨 System error occurred"
        return error_response, error_metadata, error_status


def create_advanced_gradio_interface():
    """Create advanced Gradio interface with modern design and enhanced UX"""
    
    # Advanced CSS with medical theme and animations
    css = """
    @import url('https://fonts.googleapis.com/css2?family=Love+Ya+Like+A+Sister&display=swap');

    * {
        font-family: 'Love Ya Like A Sister', cursive !important;
    }

    .gradio-container {
        background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
        min-height: 100vh;
    }
    
    .main-container {
        background: rgba(255, 255, 255, 0.95);
        backdrop-filter: blur(10px);
        border-radius: 20px;
        padding: 30px;
        margin: 20px;
        box-shadow: 0 20px 40px rgba(0,0,0,0.1);
        border: 1px solid rgba(255,255,255,0.2);
    }
    
    .header-section {
        background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
        color: white;
        border-radius: 15px;
        padding: 25px;
        margin-bottom: 25px;
        text-align: center;
        box-shadow: 0 10px 30px rgba(102, 126, 234, 0.3);
    }
    
    .query-container {
        background: linear-gradient(135deg, #f8f9ff 0%, #e8f2ff 100%);
        border-radius: 15px;
        padding: 20px;
        margin: 15px 0;
        border: 2px solid #667eea;
        transition: all 0.3s ease;
    }
    
    .query-container:hover {
        transform: translateY(-2px);
        box-shadow: 0 10px 25px rgba(102, 126, 234, 0.2);
    }
    
    .query-input {
        border: none !important;
        background: white !important;
        border-radius: 12px !important;
        padding: 15px !important;
        font-size: 16px !important;
        box-shadow: 0 4px 15px rgba(0,0,0,0.1) !important;
        transition: all 0.3s ease !important;
    }
    
    .query-input:focus {
        transform: scale(1.02) !important;
        box-shadow: 0 8px 25px rgba(102, 126, 234, 0.3) !important;
    }
    
    .response-container {
        background: linear-gradient(135deg, #fff 0%, #f8f9ff 100%);
        border-radius: 15px;
        padding: 20px;
        margin: 15px 0;
        border: 2px solid #4CAF50;
        min-height: 300px;
    }
    
    .response-output {
        border: none !important;
        background: transparent !important;
        font-size: 15px !important;
        line-height: 1.7 !important;
        color: #2c3e50 !important;
    }
    
    .metadata-container {
        background: linear-gradient(135deg, #e3f2fd 0%, #bbdefb 100%);
        border-radius: 12px;
        padding: 15px;
        margin: 10px 0;
        border-left: 5px solid #2196F3;
    }
    
    .metadata-output {
        border: none !important;
        background: transparent !important;
        font-size: 13px !important;
        color: #1565c0 !important;
        font-weight: 500 !important;
    }
    
    .status-container {
        background: linear-gradient(135deg, #e8f5e8 0%, #c8e6c9 100%);
        border-radius: 12px;
        padding: 15px;
        margin: 10px 0;
        border-left: 5px solid #4CAF50;
    }
    
    .status-output {
        border: none !important;
        background: transparent !important;
        font-size: 13px !important;
        color: #2e7d32 !important;
        font-weight: 500 !important;
    }
    
    .submit-btn {
        background: linear-gradient(135deg, #4CAF50 0%, #45a049 100%) !important;
        color: white !important;
        border: none !important;
        border-radius: 12px !important;
        padding: 15px 30px !important;
        font-size: 16px !important;
        font-weight: 600 !important;
        cursor: pointer !important;
        transition: all 0.3s ease !important;
        box-shadow: 0 6px 20px rgba(76, 175, 80, 0.3) !important;
    }
    
    .submit-btn:hover {
        transform: translateY(-3px) !important;
        box-shadow: 0 10px 30px rgba(76, 175, 80, 0.4) !important;
    }
    
    .clear-btn {
        background: linear-gradient(135deg, #ff7043 0%, #ff5722 100%) !important;
        color: white !important;
        border: none !important;
        border-radius: 12px !important;
        padding: 15px 25px !important;
        font-size: 14px !important;
        font-weight: 500 !important;
        transition: all 0.3s ease !important;
    }
    
    .clear-btn:hover {
        transform: translateY(-2px) !important;
        box-shadow: 0 8px 20px rgba(255, 87, 34, 0.3) !important;
    }
    
    .emergency-notice {
        background: linear-gradient(135deg, #ffebee 0%, #ffcdd2 100%);
        border: 2px solid #f44336;
        border-radius: 15px;
        padding: 20px;
        margin: 20px 0;
        color: #c62828;
        font-weight: 600;
        animation: pulse 2s infinite;
    }
    
    @keyframes pulse {
        0% { box-shadow: 0 0 0 0 rgba(244, 67, 54, 0.4); }
        70% { box-shadow: 0 0 0 10px rgba(244, 67, 54, 0); }
        100% { box-shadow: 0 0 0 0 rgba(244, 67, 54, 0); }
    }
    
    .gaza-context {
        background: linear-gradient(135deg, #e8f5e8 0%, #c8e6c9 100%);
        border: 2px solid #4caf50;
        border-radius: 15px;
        padding: 20px;
        margin: 20px 0;
        color: #2e7d32;
        font-weight: 500;
    }
    
    .sidebar-container {
        background: linear-gradient(135deg, #f5f7fa 0%, #c3cfe2 100%);
        border-radius: 15px;
        padding: 20px;
        margin: 10px 0;
        border: 1px solid rgba(0,0,0,0.1);
    }
    
    .example-container {
        background: white;
        border-radius: 12px;
        padding: 20px;
        margin: 15px 0;
        box-shadow: 0 4px 15px rgba(0,0,0,0.1);
    }
    
    .progress-container {
        margin: 15px 0;
        padding: 10px;
        background: rgba(255,255,255,0.8);
        border-radius: 10px;
    }
    
    .footer-section {
        background: linear-gradient(135deg, #37474f 0%, #263238 100%);
        color: white;
        border-radius: 15px;
        padding: 20px;
        margin-top: 30px;
        text-align: center;
    }

    /* GLOBAL TEXT FIXES */
.gradio-container, 
.query-container, 
.response-container, 
.metadata-container, 
.status-container {
    color: white !important;
}

.query-input, 
.response-output, 
.metadata-output, 
.status-output {
    color: white !important;
    background-color: rgba(0, 0, 0, 0.2) !important;
}

/* BANNER-INSPIRED PANEL BACKGROUNDS */
.query-container,
.response-container,
.metadata-container,
.status-container {
    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%) !important;
    border: 2px solid #ffffff22 !important;
    border-radius: 15px !important;
    box-shadow: 0 10px 30px rgba(102, 126, 234, 0.3);
}

/* EXAMPLE SECTION BUTTON STYLING */
.example-container .example {
    background: linear-gradient(135deg, #667eea 0%, #764ba2 100%) !important;
    color: white !important;
    font-weight: 600 !important;
    border-radius: 12px !important;
    padding: 15px !important;
    margin: 10px !important;
    text-align: center !important;
    box-shadow: 0 6px 20px rgba(0, 0, 0, 0.1);
    transition: all 0.3s ease;
    cursor: pointer;
}

.example-container .example:hover {
    transform: scale(1.03);
    box-shadow: 0 10px 30px rgba(102, 126, 234, 0.4);
}

/* MAKE HEADER + EXAMPLES MORE PROMINENT */
.header-section {
    color: white !important;
    text-shadow: 0px 0px 6px rgba(0,0,0,0.4);
}

.example-container {
    margin-top: -20px !important;
}
    """

    with gr.Blocks(
        css=css, 
        title="🏥 Advanced Gaza First Aid Assistant",
        theme=gr.themes.Soft(
            primary_hue="blue",
            secondary_hue="green",
            neutral_hue="slate"
        )
    ) as interface:
        
        # Header Section
        with gr.Row(elem_classes=["main-container"]):
            gr.HTML("""
            <div class="header-section">
                <h1 style="margin: 0; font-size: 2.5em; font-weight: 700;">
                    🏥 Advanced Gaza First Aid Assistant
                </h1>
                <h2 style="margin: 10px 0 0 0; font-size: 1.2em; font-weight: 400; opacity: 0.9;">
                    AI-Powered Medical Guidance for Gaza Healthcare Workers
                </h2>
                <p style="margin: 15px 0 0 0; font-size: 1em; opacity: 0.8;">
                    Enhanced with 768-dimensional medical embeddings • Advanced FAISS indexing • Real-time safety validation
                </p>
            </div>
            """)
        
        # Main Interface
        with gr.Row(elem_classes=["main-container"]):
            with gr.Column(scale=2):
                # Query Input Section
                with gr.Group(elem_classes=["query-container"]):
                    gr.Markdown("### 🩺 Medical Query Input")
                    query_input = gr.Textbox(
                        label="Describe your medical situation",
                        placeholder="Enter your first aid question or describe the medical emergency...",
                        lines=4,
                        elem_classes=["query-input"]
                    )
                    
                    with gr.Row():
                        submit_btn = gr.Button(
                            "🔍 Get Medical Guidance", 
                            variant="primary",
                            elem_classes=["submit-btn"],
                            scale=3
                        )
                        clear_btn = gr.Button(
                            "🗑️ Clear", 
                            variant="secondary",
                            elem_classes=["clear-btn"],
                            scale=1
                        )
            
            with gr.Column(scale=1):
                # Sidebar with Quick Access
                with gr.Group(elem_classes=["sidebar-container"]):
                    gr.Markdown("""
                    ### 🎯 Quick Access Guide
                    
                    **🚨 Emergency Priorities:**
                    - Severe bleeding control
                    - Burn treatment protocols  
                    - Airway management
                    - Trauma stabilization
                    - Shock prevention
                    
                    **🏥 Gaza-Specific Scenarios:**
                    - Limited water situations
                    - Power outage medical care
                    - Supply shortage alternatives
                    - Mass casualty protocols
                    - Improvised medical tools
                    
                    **📊 System Status:**
                    - ✅ Enhanced embeddings active
                    - ✅ Advanced indexing enabled
                    - ✅ Safety validation online
                    - ✅ Gaza context aware
                    """)
        
        # Response Section
        with gr.Row(elem_classes=["main-container"]):
            with gr.Column():
                # Main Response
                with gr.Group(elem_classes=["response-container"]):
                    gr.Markdown("### 🩹 Medical Guidance Response")
                    response_output = gr.Textbox(
                        label="AI Medical Guidance",
                        lines=15,
                        elem_classes=["response-output"],
                        interactive=False,
                        placeholder="Your medical guidance will appear here..."
                    )
                
                # Metadata and Status
                with gr.Row():
                    with gr.Column(scale=1):
                        with gr.Group(elem_classes=["metadata-container"]):
                            metadata_output = gr.Textbox(
                                label="📊 Response Metadata",
                                lines=2,
                                elem_classes=["metadata-output"],
                                interactive=False,
                                placeholder="Response metadata will appear here..."
                            )
                    
                    with gr.Column(scale=1):
                        with gr.Group(elem_classes=["status-container"]):
                            status_output = gr.Textbox(
                                label="🛡️ Safety Status",
                                lines=2,
                                elem_classes=["status-output"],
                                interactive=False,
                                placeholder="Safety validation status will appear here..."
                            )
        
        # Important Notices
        with gr.Row(elem_classes=["main-container"]):
            gr.HTML("""
            <div class="emergency-notice">
                <h3 style="margin: 0 0 10px 0;">🚨 CRITICAL EMERGENCY DISCLAIMER</h3>
                <p style="margin: 0; font-size: 1.1em;">
                    For life-threatening emergencies, seek immediate professional medical attention.<br>
                    📞 <strong>Gaza Emergency Contacts:</strong> Palestinian Red Crescent (101) | Civil Defense (102)
                </p>
            </div>
            """)
        
        with gr.Row(elem_classes=["main-container"]):
            gr.HTML("""
            <div class="gaza-context">
                <h3 style="margin: 0 0 10px 0;">📍 Gaza Context Awareness</h3>
                <p style="margin: 0; font-size: 1em;">
                    This advanced AI system is specifically designed for Gaza's challenging conditions including 
                    limited resources, frequent power outages, and restricted medical supply access. All guidance 
                    considers these constraints and provides practical alternatives when standard treatments are unavailable.
                </p>
            </div>
            """)
        
        # Examples Section
        with gr.Row(elem_classes=["main-container"]):
            with gr.Group(elem_classes=["example-container"]):
                gr.Markdown("### 💡 Example Medical Scenarios")
                
                example_queries = [
                    "How to treat severe burns when clean water is extremely limited?",
                    "Managing gunshot wounds with only basic household supplies",
                    "Recognizing and treating infection in wounds without antibiotics",
                    "Emergency care for children during extended power outages",
                    "Treating compound fractures without proper medical equipment",
                    "Managing diabetic emergencies when insulin is unavailable",
                    "Stopping arterial bleeding with improvised tourniquets",
                    "Recognizing and treating shock in mass casualty situations",
                    "Airway management for unconscious patients without equipment",
                    "Preventing infection in surgical wounds during siege conditions"
                ]
                
                gr.Examples(
                    examples=example_queries,
                    inputs=query_input,
                    label="Click any example to try it:",
                    examples_per_page=5
                )
        
        # Event Handlers
        submit_btn.click(
            process_medical_query_with_progress,
            inputs=query_input,
            outputs=[response_output, metadata_output, status_output],
            show_progress=True
        )
        
        query_input.submit(
            process_medical_query_with_progress,
            inputs=query_input,
            outputs=[response_output, metadata_output, status_output],
            show_progress=True
        )
        
        clear_btn.click(
            lambda: ("", "", "", ""),
            outputs=[query_input, response_output, metadata_output, status_output]
        )
        
        # Footer
        with gr.Row(elem_classes=["main-container"]):
            gr.HTML("""
            <div class="footer-section">
                <h3 style="margin: 0 0 15px 0;">🔬 Advanced Technical Features</h3>
                <div style="display: grid; grid-template-columns: repeat(auto-fit, minmax(250px, 1fr)); gap: 20px; margin-bottom: 20px;">
                    <div>
                        <strong>🧠 Enhanced AI:</strong><br>
                        768-dimensional medical embeddings<br>
                        Advanced FAISS IVF indexing<br>
                        Optimized LLM quantization
                    </div>
                    <div>
                        <strong>🛡️ Safety Systems:</strong><br>
                        Real-time medical validation<br>
                        Contraindication detection<br>
                        Gaza-specific risk assessment
                    </div>
                    <div>
                        <strong>⚡ Performance:</strong><br>
                        Async processing pipeline<br>
                        Intelligent response caching<br>
                        Progressive loading indicators
                    </div>
                </div>
                <hr style="border: 1px solid rgba(255,255,255,0.2); margin: 20px 0;">
                <p style="margin: 0; opacity: 0.8;">
                    <strong>⚕️ Medical Disclaimer:</strong> This AI assistant provides educational guidance based on established medical protocols. 
                    It is designed to support, not replace, medical professionals. Always consult qualified healthcare providers for definitive care.
                </p>
            </div>
            """)
    
    return interface

def main():
    """Enhanced main function with comprehensive error handling and system monitoring"""
    logger.info("🚀 Starting Advanced Gaza First Aid Assistant")
    
    try:
        # System initialization with detailed logging
        logger.info("🔧 Pre-initializing enhanced RAG system...")
        system = initialize_enhanced_system()
        
        # Verify system components
        logger.info("✅ Knowledge base initialized")
        logger.info("✅ Medical fact checker ready")
        logger.info("✅ Enhanced embeddings loaded")
        logger.info("✅ Advanced FAISS indexing active")
        
        # Create and launch advanced interface
        logger.info("🎨 Creating advanced Gradio interface...")
        interface = create_advanced_gradio_interface()
        
        logger.info("🌐 Launching advanced interface...")
        interface.launch(
            server_name="0.0.0.0",
            server_port=7860,
            share=False,
            max_threads=6,  # Increased for better async performance
            show_error=True,
            quiet=False,
            favicon_path=None,
            ssl_verify=False
        )
        
    except Exception as e:
        logger.error(f"❌ Failed to start Advanced Gaza First Aid Assistant: {e}")
        print(f"\n🚨 STARTUP ERROR: {e}")
        print("\n🔧 Troubleshooting Steps:")
        print("1. Check if all dependencies are installed: pip install -r requirements.txt")
        print("2. Ensure sufficient memory is available (minimum 4GB RAM recommended)")
        print("3. Verify data directory exists and contains medical documents")
        print("4. Check system logs for detailed error information")
        print("\n📞 For technical support, check the application logs above.")
        sys.exit(1)

if __name__ == "__main__":
    main()