app.py

import streamlit as st
from transformers import ViTForImageClassification, ViTImageProcessor
from PIL import Image
import torch
import time
import gc
import logging
from knowledge_base import KNOWLEDGE_BASE, DAMAGE_TYPES, validate_knowledge_base
from rag_utils import RAGSystem
import structlog
from typing import Optional, Dict, Any
from functools import lru_cache
from dynaconf import Dynaconf

# Configure settings
settings = Dynaconf(
    settings_files=['settings.yaml', '.secrets.yaml'],
    environments=True
)

# Configure logging
logging.basicConfig(level=logging.INFO)
logger = structlog.get_logger()

# Constants
MAX_FILE_SIZE = settings.get('max_file_size', 5 * 1024 * 1024)  # 5MB default
MAX_IMAGE_SIZE = settings.get('max_image_size', 1024)  # Maximum dimension default
MODEL = None
PROCESSOR = None
RAG_SYSTEM = None

def handle_exceptions(func):
    """Decorator for exception handling"""
    def wrapper(*args, **kwargs):
        try:
            return func(*args, **kwargs)
        except Exception as e:
            cleanup_memory()
            st.error(f"Error in {func.__name__}: {str(e)}")
            logger.error(f"Error in {func.__name__}: {str(e)}", exc_info=True)
            return None
    return wrapper

def cleanup_memory():
    """Clean up memory and GPU cache"""
    gc.collect()
    if torch.cuda.is_available():
        torch.cuda.empty_cache()

def init_session_state():
    """Initialize session state variables"""
    if 'history' not in st.session_state:
        st.session_state.history = []
    if 'dark_mode' not in st.session_state:
        st.session_state.dark_mode = False
    if 'analysis_count' not in st.session_state:
        st.session_state.analysis_count = 0

@st.cache_resource(show_spinner="Loading AI model...", ttl=3600*24)
def load_model():
    """Load and cache the model with daily refresh"""
    try:
        model_name = "google/vit-base-patch16-224"
        processor = ViTImageProcessor.from_pretrained(model_name)
        device = "cuda" if torch.cuda.is_available() else "cpu"
        model = ViTForImageClassification.from_pretrained(
            model_name,
            num_labels=len(DAMAGE_TYPES),
            ignore_mismatched_sizes=True,
        ).to(device)
        model.eval()
        logger.info("Model loaded successfully", device=device)
        return model, processor
    except Exception as e:
        logger.error("Error loading model", error=str(e))
        return None, None

def validate_upload(file) -> bool:
    """Validate uploaded file for security"""
    if not file:
        return False
    
    allowed_extensions = {'jpg', 'jpeg', 'png'}
    if not file.name.lower().endswith(tuple(allowed_extensions)):
        st.error("Invalid file type. Please upload a JPG or PNG image.")
        return False
    
    if file.size > MAX_FILE_SIZE:
        st.error(f"File too large. Maximum size is {MAX_FILE_SIZE/1024/1024:.1f}MB.")
        return False
    
    if file.type not in ['image/jpeg', 'image/png']:
        st.error("Invalid file content type.")
        return False
    
    return True

@handle_exceptions
def preprocess_image(uploaded_file) -> Optional[Image.Image]:
    """Preprocess and validate uploaded image"""
    image = Image.open(uploaded_file)
    if max(image.size) > MAX_IMAGE_SIZE:
        ratio = MAX_IMAGE_SIZE / max(image.size)
        new_size = tuple([int(dim * ratio) for dim in image.size])
        image = image.resize(new_size, Image.Resampling.LANCZOS)
    return image

@handle_exceptions
def analyze_damage(image: Image.Image, model: ViTForImageClassification, 
                  processor: ViTImageProcessor) -> Optional[torch.Tensor]:
    """Analyze structural damage in the image"""
    progress_bar = st.progress(0)
    stages = ['Preprocessing', 'Analysis', 'Results Generation']
    
    try:
        device = next(model.parameters()).device
        
        for i, stage in enumerate(stages):
            progress_bar.progress((i + 1) / len(stages))
            st.write(f"Stage {i+1}/{len(stages)}: {stage}")
            
            if i == 0:  # Preprocessing
                image = image.convert('RGB')
                inputs = processor(images=image, return_tensors="pt")
                inputs = {k: v.to(device) for k, v in inputs.items()}
            
            elif i == 1:  # Analysis
                with torch.no_grad():
                    outputs = model(**inputs)
                    probs = torch.nn.functional.softmax(outputs.logits, dim=1)[0]
            
            elif i == 2:  # Results Generation
                result = probs.cpu()
                
        return result
        
    except RuntimeError as e:
        if "out of memory" in str(e):
            cleanup_memory()
            st.error("Out of memory. Please try with a smaller image.")
        else:
            st.error(f"Error analyzing image: {str(e)}")
        return None

def generate_downloadable_report(analysis_results: Dict):
    """Generate a downloadable PDF report"""
    try:
        import io
        from reportlab.lib import colors
        from reportlab.lib.pagesizes import letter
        from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer
        from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle

        buffer = io.BytesIO()
        doc = SimpleDocTemplate(buffer, pagesize=letter)
        styles = getSampleStyleSheet()
        story = []

        # Add title
        story.append(Paragraph("Structural Damage Analysis Report", styles['Title']))
        story.append(Spacer(1, 12))

        # Add analysis results
        for damage_type, details in analysis_results.items():
            story.append(Paragraph(f"Damage Type: {damage_type}", styles['Heading1']))
            story.append(Paragraph(f"Confidence: {details['confidence']}%", styles['Normal']))
            story.append(Paragraph("Recommendations:", styles['Heading2']))
            for rec in details['recommendations']:
                story.append(Paragraph(f"• {rec}", styles['Normal']))
            story.append(Spacer(1, 12))

        doc.build(story)
        pdf = buffer.getvalue()
        buffer.close()
        return pdf
    except Exception as e:
        logger.error(f"Error generating report: {str(e)}")
        return None

def display_analysis_results(predictions: torch.Tensor, analysis_time: float):
    """Display analysis results with damage details"""
    st.markdown("### 📊 Analysis Results")
    st.markdown(f"*Analysis completed in {analysis_time:.2f} seconds*")
    
    analysis_results = {}
    detected = False
    
    for idx, prob in enumerate(predictions):
        confidence = float(prob) * 100
        if confidence > 15:
            detected = True
            damage_type = DAMAGE_TYPES[idx]['name']
            
            with st.expander(f"{damage_type.replace('_', ' ').title()} - {confidence:.1f}%", expanded=True):
                st.progress(confidence / 100)
                
                # Get enhanced analysis from RAG system
                analysis = RAG_SYSTEM.get_enhanced_analysis(damage_type, confidence)
                
                tabs = st.tabs(["📋 Details", "🔧 Repairs", "⚠️ Safety"])
                
                with tabs[0]:
                    for detail in analysis['technical_details']:
                        st.markdown(detail)
                
                with tabs[1]:
                    for rec in analysis['expert_recommendations']:
                        st.markdown(rec)
                
                with tabs[2]:
                    for safety in analysis['safety_considerations']:
                        st.warning(safety)
                
                analysis_results[damage_type] = {
                    'confidence': confidence,
                    'recommendations': analysis['expert_recommendations']
                }
    
    if not detected:
        st.info("No significant structural damage detected. Regular maintenance recommended.")
    else:
        # Generate download button for report
        pdf_report = generate_downloadable_report(analysis_results)
        if pdf_report:
            st.download_button(
                label="Download Analysis Report",
                data=pdf_report,
                file_name="damage_analysis_report.pdf",
                mime="application/pdf"
            )

def main():
    """Main application function"""
    st.set_page_config(
        page_title="Structural Damage Analyzer Pro",
        page_icon="🏗️",
        layout="wide",
        initial_sidebar_state="expanded"
    )
    
    init_session_state()
    
    st.markdown("""
        <div style='text-align: center; padding: 1rem;'>
            <h1>🏗️ Structural Damage Analyzer Pro</h1>
            <p style='font-size: 1.2rem;'>Advanced AI-powered structural damage assessment tool</p>
        </div>
    """, unsafe_allow_html=True)
    
    # Sidebar
    with st.sidebar:
        st.markdown("### ⚙️ Settings")
        st.session_state.dark_mode = st.toggle("Dark Mode", st.session_state.dark_mode)
        st.markdown("### 📖 Analysis History")
        if st.session_state.history:
            for item in st.session_state.history[-5:]:
                st.markdown(f"- {item}")
    
    # Load model and initialize RAG system
    global MODEL, PROCESSOR, RAG_SYSTEM
    if MODEL is None or PROCESSOR is None:
        MODEL, PROCESSOR = load_model()
        if MODEL is None:
            st.error("Failed to load model. Please refresh the page.")
            return
    
    if RAG_SYSTEM is None:
        RAG_SYSTEM = RAGSystem()
        RAG_SYSTEM.initialize_knowledge_base(KNOWLEDGE_BASE)
    
    # Validate knowledge base
    if not validate_knowledge_base():
        st.error("Knowledge base validation failed. Please check the logs.")
        return
    
    # File upload
    uploaded_file = st.file_uploader(
        "Upload an image for analysis",
        type=['jpg', 'jpeg', 'png'],
        help="Supported formats: JPG, JPEG, PNG"
    )
    
    if uploaded_file and validate_upload(uploaded_file):
        try:
            image = preprocess_image(uploaded_file)
            if image is None:
                return
            
            col1, col2 = st.columns([1, 1])
            
            with col1:
                st.image(image, caption="Uploaded Structure", use_column_width=True)
            
            with col2:
                start_time = time.time()
                predictions = analyze_damage(image, MODEL, PROCESSOR)
                
                if predictions is not None:
                    analysis_time = time.time() - start_time
                    display_analysis_results(predictions, analysis_time)
                    st.session_state.history.append(f"Analyzed {uploaded_file.name}")
                    st.session_state.analysis_count += 1
        
        except Exception as e:
            logger.error("Error in main processing loop", error=str(e))
            cleanup_memory()
            st.error("An error occurred during processing. Please try again.")
    
    # Footer
    st.markdown("---")
    st.markdown("""
        <div style='text-align: center'>
            <p>🏗️ Structural Damage Analyzer Pro | Built with Streamlit & Transformers</p>
            <p style='font-size: 0.8rem;'>For professional use only. Always consult with a structural engineer.</p>
        </div>
    """, unsafe_allow_html=True)

if __name__ == "__main__":
    main()