app.py

import streamlit as st
from PIL import Image, ImageFile
import sys
import os
import json
from datetime import datetime
import logging
import tempfile
import io

st.set_page_config(
    page_title="Device Defect Diagnosis System",
    layout="wide",
    initial_sidebar_state="expanded"
)
# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    handlers=[
        logging.StreamHandler(sys.stdout)
    ]
)

logger = logging.getLogger(__name__)

# Set page config first (must be first Streamlit command)

if "booted" not in st.session_state:
    logger.info("DEVICE DEFECT DIAGNOSIS SYSTEM - STARTING")
    st.session_state.booted = True


HF_USERNAME = "palakmathur"
# sys.path.append(os.path.join(os.path.dirname(__file__), '..', 'src'))

# Import only huggingface_hub at top level (needed for model downloads)
try:
    from huggingface_hub import hf_hub_download
    if "imports_logged" not in st.session_state:
        logger.info("✓ huggingface_hub imported successfully")
        st.session_state.imports_logged = True
except ImportError as e:
    logger.error(f" Failed to import huggingface_hub: {e}")
    st.error("huggingface_hub not installed")
    st.stop()

# Heavy model imports moved inside load_system() to avoid re-importing on every rerun

ImageFile.LOAD_TRUNCATED_IMAGES = True

st.markdown("""
<style>
    /* Hide default Streamlit header */
    header[data-testid="stHeader"] {
        display: none;
    }
    
    .main-header {
        font-size: 2.5rem;
        font-weight: bold;
        color: #1f77b4;
        text-align: center;
        padding: 1rem;
    }
    .sub-header {
        font-size: 1.2rem;
        color: #666;
        text-align: center;
        margin-bottom: 2rem;
    }
    .metric-card {
        background: linear-gradient(135deg, #667eea 0%, #764ba2 100%);
        padding: 1.5rem;
        border-radius: 10px;
        color: white;
        margin: 0.5rem 0;
    }
    .success-box {
        background: linear-gradient(135deg, #1e7e34 0%, #28a745 100%);
        border-left: 5px solid #4cff4c;
        padding: 1rem;
        margin: 1rem 0;
        border-radius: 5px;
        color: white;
        font-weight: 500;
        box-shadow: 0 2px 8px rgba(40, 167, 69, 0.3);
    }
    .warning-box {
        background: linear-gradient(135deg, #c43a00 0%, #dc3545 100%);
        border-left: 5px solid #ffcc00;
        padding: 1rem;
        margin: 1rem 0;
        border-radius: 5px;
        color: white;
        font-weight: 500;
        box-shadow: 0 2px 8px rgba(220, 53, 69, 0.3);
    }
    .error-box {
        background-color: #f8d7da;
        border-left: 5px solid #dc3545;
        padding: 1rem;
        margin: 1rem 0;
        border-radius: 5px;
    }
    .defect-name {
        font-size: 1.5rem;
        font-weight: bold;
        color: white;
        padding: 0.5rem 0;
        margin: 0.5rem 0;
    }
    .defect-info {
        background: linear-gradient(135deg, #1a5490 0%, #1f77b4 100%);
        padding: 1.2rem;
        border-radius: 8px;
        border-left: 4px solid #5dade2;
        margin: 1rem 0;
        box-shadow: 0 2px 8px rgba(31, 119, 180, 0.3);
    }
</style>
""", unsafe_allow_html=True)


@st.cache_resource(show_spinner=False)
def load_system():
    """Load the diagnosis system (cached) - returns immediately with status"""
    status = {"loaded": False, "error": None}
    
    logger.info("STARTING SYSTEM INITIALIZATION")
    
    
    # Import heavy modules inside function to avoid re-importing on Streamlit reruns
    try:
        from domain_validator import DomainValidator
        from description_extractor import DescriptionExtractor
        from defect_matcher import DefectMatcher
        from condition_grader import ConditionGrader
        from predictor.price_predictor import PricePredictor
        logger.info(" All custom modules imported successfully")
    except ImportError as e:
        logger.error(f"Import error: {e}", exc_info=True)
        status["error"] = f"Import error: {e}"
        return None, None, None, None, None, status
    
    try:
        # Quick components
        logger.info("Loading DomainValidator...")
        validator = DomainValidator()
        logger.info("DomainValidator loaded successfully")
        
        logger.info("Loading DescriptionExtractor...")
        extractor = DescriptionExtractor()
        logger.info(" DescriptionExtractor loaded successfully")
        
        logger.info("Loading ConditionGrader...")
        grader = ConditionGrader()
        logger.info(" ConditionGrader loaded successfully")
        
        # Heavy model loading with timeout handling
        matcher = None
        predictor = None
        
        try:
            # Load CLIP model
            logger.info("Loading DefectMatcher with fine-tuned model from HuggingFace...")
            logger.info("Model repo: palakmathur/device-defect-clip")
            matcher = DefectMatcher(use_finetuned=True)
            #matcher = DefectMatcher(use_finetuned=False)

            status["clip_loaded"] = True
            logger.info(" Fine-tuned CLIP model loaded successfully")
        except Exception as e:
            logger.error(f" ERROR loading fine-tuned CLIP model: {e}", exc_info=True)
            status["clip_error"] = str(e)[:200]
            try:
                logger.info("Attempting to load base CLIP model as fallback...")
                matcher = DefectMatcher(use_finetuned=False)
                status["clip_loaded"] = "fallback"
                logger.info(" Base CLIP model loaded successfully")
            except Exception as e2:
                logger.error(f" ERROR loading base CLIP model: {e2}", exc_info=True)
                status["clip_error"] = str(e2)[:200]
        
        try:
            # Load price predictor
            logger.info("Loading PricePredictor...")
            predictor = PricePredictor()
            logger.info("Downloading price prediction model from HuggingFace...")
            logger.info("Model repo: palakmathur/device-defect-pricing")
            predictor.load_model()
            status["price_loaded"] = True
            logger.info(" Price prediction model loaded successfully")
        except Exception as e:
            logger.error(f" ERROR loading price predictor: {e}", exc_info=True)
            status["price_error"] = str(e)[:200]
            predictor = None
        
        status["loaded"] = True

        logger.info("SYSTEM INITIALIZATION COMPLETE!")

        return validator, extractor, matcher, grader, predictor, status
        
    except Exception as e:
        logger.error(f" CRITICAL ERROR during system initialization: {e}", exc_info=True)
        status["error"] = str(e)
        return None, None, None, None, None, status

# Main application flow
if "app_started" not in st.session_state:
    logger.info("Main application started")
    st.session_state.app_started = True
    
st.markdown('<div class="main-header"> Device Defect Diagnosis System</div>', unsafe_allow_html=True)
st.markdown('<div class="sub-header">Upload a device image to detect defects and estimate resale value</div>', unsafe_allow_html=True)

if "system" not in st.session_state:
    st.session_state.system = None

if "loading" not in st.session_state:
    st.session_state.loading = False

if "processing" not in st.session_state:
    st.session_state.processing = False

# Show loading status - distinguish between "not yet loaded" vs "failed to load"
if st.session_state.system is None:
    st.info(" Click **Load  Models** below to begin.")
    st.warning("Models not loaded yet.")
    
    if st.button(" Load  Models"):
        with st.spinner("Loading models (first time only)..."):
            st.session_state.system = load_system()
            st.session_state.loading = False
    st.stop()

# Models are loaded - extract from session state
validator, extractor, matcher, grader, predictor, status = st.session_state.system

logger.info(f"System load status: {status}")

# Check if loading failed
if not status.get("loaded"):
    error_msg = status.get('error', 'Unknown error')
    st.error(f" Model loading failed: {error_msg}")
    st.stop()

# Show success message now that models are confirmed loaded
st.success(" System initialized successfully! All models loaded.")

# Show model status
with st.expander(" System Status", expanded=False):
    if status.get("clip_loaded") == True:
        st.success(" Fine-tuned CLIP model loaded")
    elif status.get("clip_loaded") == "fallback":
        st.warning(" Using base CLIP model (fine-tuned model unavailable)")
    else:
        st.error(f" CLIP model error: {status.get('clip_error', 'Unknown')}")
    
    if status.get("price_loaded"):
        st.success(" Price prediction model loaded")
    else:
        st.warning(f" Price prediction unavailable: {status.get('price_error', 'Not loaded')}")

# Check if system loaded successfully
if validator is None or matcher is None:
    st.error(" Critical components failed to load. Please try refreshing the page.")
    st.stop()

with st.sidebar:
    st.header("Instructions:-")
    st.markdown("""
    1. **Upload** a device image (phone/laptop)
    2. **Describe** the issue (optional)
    3. **Enter** device details for pricing
    4. **Click** Diagnose to get results
    
    ---
    
    ###  Supported Devices
    -  Smartphones
    -  Laptops
    -  Desktop computers
    
    ---
    
    ###  Understanding Results
    
    **Confidence Score**  
    How certain the model is about its detection (0-100%)
    - 90%+ : High confidence
    - 70-90% : Good confidence
    - <70% : Review needed
    
    **Severity Score** (0-10)  
    Impact of the defect on device functionality
    - 8-10 : Critical (needs repair)
    - 5-7 : Moderate (affects usage)
    - 0-4 : Minor (cosmetic)
    
    **Condition Score** (0-10)  
    Overall device condition after inspection
    - 9-10 : Excellent (Grade A)
    - 7-9 : Good (Grade B)
    - 5-7 : Fair (Grade C)
    - 3-5 : Poor (Grade D)
    - 0-3 : Bad (Grade F)
    
    ---
    
    ###  Detected Issues:-
    - Screen damage
    - Physical defects
    - Component failures
    - And more...
    """)
    
    st.markdown("---")
    st.markdown("### System Stats :-")
    st.metric("Models Loaded", "5/5")
    st.metric("Defect Types", "15+")
    st.metric("Avg Response Time", "8.2s")

col1, col2 = st.columns([1, 1])

with col1:
    st.header(" Upload & Configure")
    
    #
    uploaded_file = st.file_uploader(
        "Choose a device image",
        type=['jpg', 'jpeg', 'png'],
        help="Upload a clear image of your device",
        key="uploader"
    )
        
    if uploaded_file:
        st.session_state.image_bytes = uploaded_file.getvalue()
        st.session_state.image_name = uploaded_file.name

        
    # Load image from session state (persists across reruns)
    image = None
    if "image_bytes" in st.session_state:
        image = Image.open(io.BytesIO(st.session_state.image_bytes))
        st.image(image, caption="Uploaded Image", use_column_width=False)
        
    description = st.text_area(
        "Describe the issue (optional)",
        placeholder="e.g., My phone screen cracked after I dropped it. Touch is not working properly.",
        height=100,
        help="Providing a description improves accuracy"
    )
        
    st.subheader(" Device Information (for price prediction)")
        
    col_a, col_b = st.columns(2)
        
    with col_a:
        brand = st.selectbox(
            "Brand",
            ["Apple", "Samsung", "OnePlus", "Dell", "HP", "Lenovo", "Other"]
        )
            
        device_type = st.selectbox(
            "Device Type",
            ["Phone", "Laptop"]
        )
        
    model_options = {
        "Apple": {
            "Phone": ["iPhone 15 Pro Max", "iPhone 15 Pro", "iPhone 15", "iPhone 14 Pro Max", "iPhone 14 Pro", "iPhone 14", "iPhone 13 Pro Max", "iPhone 13 Pro", "iPhone 13", "iPhone 12 Pro Max", "iPhone 12 Pro", "iPhone 12", "iPhone 11 Pro Max", "iPhone 11 Pro", "iPhone 11", "iPhone XS Max", "iPhone XS", "iPhone XR", "iPhone X", "iPhone 8 Plus", "iPhone 8"],
            "Laptop": ["MacBook Pro 16\" M3 Max", "MacBook Pro 14\" M3 Pro", "MacBook Pro 16\" M2 Max", "MacBook Pro 14\" M2 Pro", "MacBook Air M3", "MacBook Air M2", "MacBook Air M1", "MacBook Pro 13\" M1"]
        },
        "Samsung": {
            "Phone": ["Galaxy S24 Ultra", "Galaxy S24+", "Galaxy S24", "Galaxy S23 Ultra", "Galaxy S23+", "Galaxy S23", "Galaxy S22 Ultra", "Galaxy S22+", "Galaxy S22", "Galaxy Z Fold 5", "Galaxy Z Flip 5", "Galaxy Z Fold 4", "Galaxy Z Flip 4", "Galaxy A54", "Galaxy A34"],
            "Laptop": ["Galaxy Book4 Pro", "Galaxy Book3 Ultra", "Galaxy Book3 Pro 360", "Galaxy Book2 Pro", "Galaxy Book2"]
        },
        "OnePlus": {
            "Phone": ["OnePlus 12", "OnePlus 11", "OnePlus 10 Pro", "OnePlus 9 Pro", "OnePlus 9", "OnePlus 8T", "OnePlus Nord 3", "OnePlus Nord 2"],
            "Laptop": []
            },
        "Dell": {
            "Phone": [],
            "Laptop": ["XPS 15", "XPS 13", "XPS 17", "Inspiron 15", "Inspiron 14", "Latitude 5430", "Latitude 7430", "Alienware m15", "Alienware x15", "Vostro 15"]
        },
        "HP": {
            "Phone": [],
            "Laptop": ["Pavilion 15", "Pavilion 14", "Envy 15", "Envy 13", "Spectre x360 14", "Spectre x360 16", "EliteBook 840", "EliteBook 850", "ProBook 450", "Omen 15"]
        },
        "Lenovo": {
            "Phone": [],
            "Laptop": ["ThinkPad X1 Carbon", "ThinkPad X1 Yoga", "ThinkPad T14", "ThinkPad T16", "IdeaPad Slim 5", "IdeaPad Gaming 3", "Yoga 9i", "Yoga 7i", "Legion 5 Pro"]
        },
        "Other": {
            "Phone": ["Other Model"],
            "Laptop": ["Other Model"]
        }
    }
        
    available_models = model_options.get(brand, {}).get(device_type, ["Other Model"])
        
    with col_b:
        if available_models:
            model = st.selectbox("Model", available_models)
        else:
            st.warning(f"No {device_type} models available for {brand}")
            model = st.text_input("Model (enter manually)", "")
            
        original_price = st.number_input("Original Price (₹)", min_value=0, value=79900, step=1000)
        
    age_months = st.slider("Age (months)", 0, 60, 18, help="How old is the device?")
    if st.session_state.processing:
        st.warning("⏳ Processing… please wait")
        # Form submit button
    with st.form("diagnosis_form"):
        submitted = st.form_submit_button(" Diagnose Device", type="primary")

with col2:
    st.header("Diagnosis Results :- ")
    
    if submitted and "image_bytes" in st.session_state:
            
            # Create unique temporary file for this user
            with tempfile.NamedTemporaryFile(delete=False, suffix=".jpg") as tmp:
                temp_path = tmp.name
            
            # HARD safety processing
            image = Image.open(io.BytesIO(st.session_state.image_bytes)).convert("RGB")
            image.thumbnail((1024, 1024))
            image.save(temp_path, "JPEG", quality=85, optimize=True)
            
            progress_bar = st.progress(0)
            status_text = st.empty()
            
            # Mark as processing to prevent double-click
            st.session_state.processing = True
            
            try:
                status_text.text("Stage 1/5: Validating device...")
                progress_bar.progress(20)
                
                validation = validator.validate(temp_path)
                
                if not validation['is_valid']:
                    st.markdown(f'<div class="error-box"> <b>Invalid Image</b><br>{validation["reason"]}</div>', unsafe_allow_html=True)
                    st.stop()
                
                status_text.text("Stage 2/5: Processing description...")
                progress_bar.progress(40)
                    
                if description:
                    desc_info = extractor.extract(description)
                    search_text = extractor.create_search_text(desc_info)
                else:
                    search_text = None
                    
                status_text.text(" Stage 3/5: Detecting defects...")
                progress_bar.progress(60)
                    
                match_result = matcher.match(temp_path, search_text, top_k=3)
                defect = match_result['top_match']
                    
                status_text.text("Stage 4/5: Grading condition...")
                progress_bar.progress(80)
                    
                grading = grader.assign_grade([defect])
                    
                status_text.text("Stage 5/5: Predicting price...")
                progress_bar.progress(90)
                    
                price_result = None
                if predictor:
                    try:
                        device_info = {
                            'brand': brand,
                            'model': model,
                            'original_price': original_price,
                            'age_months': age_months
                        }
                        price_result = predictor.predict(device_info, [defect], grading['condition_score'])
                    except:
                        pass
                    
                progress_bar.progress(100)
                status_text.text(" Diagnosis complete!!!")
                    
                st.markdown("---")
                    
                st.markdown(f'<div class="success-box"> <b>Valid {validation["device_type"].title()} Detected</b> (Confidence: {validation["confidence"]:.1%})</div>', unsafe_allow_html=True)
                    
                st.subheader(" Detected Issue")
                    
                severity_color = "#dc3545" if defect.get('critical', False) else "#ffc107" if defect['severity_score'] > 6 else "#28a745"
                    
                st.markdown(f'<div class="defect-info"><div class="defect-name"> {defect["name"]}</div></div>', unsafe_allow_html=True)
                    
                col_r1, col_r2 = st.columns(2)
                col_r1.metric(
                        "Confidence", 
                        f"{match_result['confidence']:.1%}",
                        help="How certain the AI is about this detection. Higher is better."
                    )
                col_r2.metric(
                        "Severity", 
                        f"{defect['severity_score']}/10",
                        help="Impact of this defect: 0-4 (Minor), 5-7 (Moderate), 8-10 (Critical)"
                    )
                    
                if defect['severity_score'] >= 8:
                    severity_msg = " **High Severity:** This defect significantly impacts device functionality and requires immediate attention."
                elif defect['severity_score'] >= 5:
                    severity_msg = " **Moderate Severity:** This defect affects device usage. Consider repair before resale."
                else:
                    severity_msg = " **Low Severity:** This is mostly a cosmetic issue with minimal impact on functionality."
                
                st.info(severity_msg)
                
                if defect.get('critical', False):
                    st.markdown(f'<div class="warning-box"> <b>Critical Issue Detected</b><br>Immediate attention required</div>', unsafe_allow_html=True)
                    
                st.subheader(" Condition Assessment")
                    
                col_c1, col_c2 = st.columns(2)
                col_c1.metric(
                        "Grade", 
                        grading['grade'], 
                        help="Letter grade from A (Excellent) to F (Bad) based on overall condition"
                    )
                col_c2.metric(
                        "Condition Score", 
                        f"{grading['condition_score']}/10",
                        help="Numerical score: 10 is perfect, 0 is non-functional. Affects resale value directly."
                    )
                    
                st.caption(f"**{grading['description']}**")
                    
                with st.expander("What does this grade mean???", expanded=False):
                    grade_info = {
                        'A': "**Excellent Condition** - Device is like new with minimal to no defects. Commands premium resale price.",
                        'B': "**Good Condition** - Device shows minor wear but fully functional. Good resale value.",
                        'C': "**Fair Condition** - Device has noticeable defects but still usable. Moderate resale value.",
                        'D': "**Poor Condition** - Device has significant issues affecting usability. Low resale value.",
                        'F': "**Bad Condition** - Device has critical defects or is barely functional. Very low resale value."
                    }
                    st.write(grade_info.get(grading['grade'], "Assessment completed"))
                    
                    st.write("""\n**How it's calculated:**
                    - Based on severity and number of defects
                    - Physical damage has higher impact
                    - Critical defects significantly lower the grade
                    - Overall device age and wear considered""")
                
                if price_result:
                    st.subheader("Resale Valuation")
                    
                    col_p1, col_p2, col_p3 = st.columns(3)
                    col_p1.metric(
                        "Estimated Price", 
                        f"₹{price_result['predicted_price']:,}",
                        help="AI-predicted resale price based on condition, defects, and market data"
                    )
                    col_p2.metric(
                        "Min Price", 
                        f"₹{price_result['price_range']['min']:,}",
                        help="Minimum expected price in current market conditions"
                    )
                    col_p3.metric(
                        "Max Price", 
                        f"₹{price_result['price_range']['max']:,}",
                        help="Maximum expected price for this condition"
                    )
                    
                    st.caption(f"**Prediction Confidence:** {price_result['confidence']:.0%} - How reliable this price estimate is based on available data")
                    
                    with st.expander(" Price Impact Factors"):
                        for factor in price_result['depreciation_factors']:
                            st.markdown(f"• **{factor['factor']}**: {factor['impact']} - {factor['description']}")
                
                if len(match_result['all_matches']) > 1:
                    with st.expander(" Alternative Diagnoses"):
                        for i, alt in enumerate(match_result['all_matches'][1:3], 2):
                            st.markdown(f"{i}. **{alt['defect']['name']}** - Confidence: {alt['confidence']:.1%}")
                    
                st.markdown("---")
                
                # Generate PDF report
                try:
                    with st.spinner("Generating report..."):
                        from fpdf import FPDF
                        from fpdf.enums import XPos, YPos
                        
                        class DiagnosisReport(FPDF):
                            def header(self):
                                self.set_font('Helvetica', 'B', 20)
                                self.set_text_color(31, 119, 180)
                                self.cell(0, 10, 'Device Diagnosis Report', new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='C')
                                self.set_font('Helvetica', '', 10)
                                self.set_text_color(100, 100, 100)
                                self.cell(0, 6, f"Generated: {datetime.now().strftime('%B %d, %Y at %I:%M %p')}", new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='C')
                                self.ln(10)
                            
                            def footer(self):
                                self.set_y(-15)
                                self.set_font('Helvetica', 'I', 8)
                                self.set_text_color(128, 128, 128)
                        
                        pdf = DiagnosisReport()
                        pdf.add_page()
                        pdf.set_auto_page_break(auto=True, margin=15)
                        
                        pdf.set_font('Helvetica', 'B', 14)
                        pdf.set_text_color(44, 62, 80)
                        pdf.cell(0, 10, 'Device Information', new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                        pdf.ln(2)
                        
                        pdf.set_font('Helvetica', '', 11)
                        pdf.set_text_color(0, 0, 0)
                        device_info = [
                            ('Device Type:', validation['device_type'].title()),
                            ('Brand:', brand),
                            ('Model:', model),
                            ('Age:', f"{age_months} months"),
                            ('Detection Confidence:', f"{validation['confidence']:.1%}")
                        ]
                        for label, value in device_info:
                            pdf.set_font('Helvetica', 'B', 11)
                            pdf.cell(60, 8, label, border=1, new_x=XPos.RIGHT, new_y=YPos.TOP, align='L')
                            pdf.set_font('Helvetica', '', 11)
                            pdf.cell(0, 8, str(value), border=1, new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                        
                        pdf.ln(8)
                        
                        pdf.set_font('Helvetica', 'B', 14)
                        pdf.set_text_color(44, 62, 80)
                        pdf.cell(0, 10, 'Detected Issue', new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                        pdf.ln(2)
                        
                        pdf.set_font('Helvetica', '', 11)
                        pdf.set_text_color(0, 0, 0)
                        defect_info = [
                            ('Defect Name:', defect['name']),
                            ('Confidence:', f"{match_result['confidence']:.1%}"),
                            ('Severity Score:', f"{defect['severity_score']}/10"),
                            ('Critical:', 'Yes' if defect.get('critical', False) else 'No')
                        ]
                        for label, value in defect_info:
                            pdf.set_font('Helvetica', 'B', 11)
                            pdf.cell(60, 8, label, border=1, new_x=XPos.RIGHT, new_y=YPos.TOP, align='L')
                            pdf.set_font('Helvetica', '', 11)
                            pdf.cell(0, 8, str(value), border=1, new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                        
                        pdf.ln(8)
                        
                        pdf.set_font('Helvetica', 'B', 14)
                        pdf.set_text_color(44, 62, 80)
                        pdf.cell(0, 10, 'Condition Assessment', new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                        pdf.ln(2)
                        
                        pdf.set_font('Helvetica', '', 11)
                        pdf.set_text_color(0, 0, 0)
                        condition_info = [
                            ('Grade:', grading['grade']),
                            ('Condition Score:', f"{grading['condition_score']}/10"),
                            ('Description:', grading['description'])
                        ]
                        for label, value in condition_info:
                            pdf.set_font('Helvetica', 'B', 11)
                            pdf.cell(60, 8, label, border=1, new_x=XPos.RIGHT, new_y=YPos.TOP, align='L')
                            pdf.set_font('Helvetica', '', 11)
                            pdf.cell(0, 8, str(value), border=1, new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                        
                        pdf.ln(8)
                        
                        if price_result:
                            pdf.set_font('Helvetica', 'B', 14)
                            pdf.set_text_color(44, 62, 80)
                            pdf.cell(0, 10, 'Resale Valuation', new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                            pdf.ln(2)
                            
                            pdf.set_font('Helvetica', '', 11)
                            pdf.set_text_color(0, 0, 0)
                            pricing_info = [
                                ('Estimated Price:', f"Rs. {price_result['predicted_price']:,}"),
                                ('Price Range:', f"Rs. {price_result['price_range']['min']:,} - Rs. {price_result['price_range']['max']:,}"),
                                ('Prediction Confidence:', f"{price_result['confidence']:.0%}")
                            ]
                            for label, value in pricing_info:
                                pdf.set_font('Helvetica', 'B', 11)
                                pdf.cell(60, 8, label, border=1, new_x=XPos.RIGHT, new_y=YPos.TOP, align='L')
                                pdf.set_font('Helvetica', '', 11)
                                pdf.cell(0, 8, str(value), border=1, new_x=XPos.LMARGIN, new_y=YPos.NEXT, align='L')
                        
                        pdf_data = bytes(pdf.output())
                        
                        st.download_button(
                            " Download Full Report (PDF)",
                            data=pdf_data,
                            file_name=f"diagnosis_report_{datetime.now().strftime('%Y%m%d_%H%M%S')}.pdf",
                            mime="application/pdf"
                        )
                except (ImportError, Exception) as e:
                    st.warning(f" PDF generation failed: {str(e)}")
                    report = {
                        'device': {
                            'type': validation['device_type'],
                            'brand': brand,
                            'model': model,
                            'age_months': age_months
                        },
                        'defect': {
                            'name': defect['name'],
                            'confidence': match_result['confidence'],
                            'severity': defect['severity_score']
                        },
                        'condition': {
                            'grade': grading['grade'],
                            'score': grading['condition_score']
                        }
                    }
                    
                    if price_result:
                        report['pricing'] = {
                            'estimated_price': price_result['predicted_price'],
                            'price_range': price_result['price_range']
                        }
                    
                    st.download_button(
                        " Download Full Report (JSON)",
                        data=json.dumps(report, indent=2),
                        file_name="diagnosis_report.json",
                        mime="application/json"
                    )
            
            except Exception as e:
                st.error(f" Error during diagnosis: {str(e)}")
            finally:
                # Always clean up temp file
                if os.path.exists(temp_path):
                    os.remove(temp_path)
                # Reset processing flag
                st.session_state.processing = False
    else:
      if not submitted:
        st.info(" Upload an image and click Diagnose to begin")
st.markdown("---")
st.markdown("""
<div style='text-align: center; color: #666; padding: 2rem;'>
    <p><b>Device Defect Diagnosis System</b></p>
    <p>Built with CLIP, BERT, and XGBoost</p>
</div>
""", unsafe_allow_html=True)