app.py

import gradio as gr
import pandas as pd
import json
import os
import re
from PyPDF2 import PdfReader
from collections import defaultdict
from typing import Dict, List, Optional, Tuple, Union
import html
from pathlib import Path
import fitz  # PyMuPDF
import pytesseract
from PIL import Image
import io
import secrets
import string
from huggingface_hub import HfApi, HfFolder
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
import time
import logging
import asyncio
from functools import lru_cache
import hashlib
from concurrent.futures import ThreadPoolExecutor

# ========== CONFIGURATION ==========
PROFILES_DIR = "student_profiles"
ALLOWED_FILE_TYPES = [".pdf", ".png", ".jpg", ".jpeg"]
MAX_FILE_SIZE_MB = 5
MIN_AGE = 5
MAX_AGE = 120
SESSION_TOKEN_LENGTH = 32
HF_TOKEN = os.getenv("HF_TOKEN")
SESSION_TIMEOUT = 3600  # 1 hour session timeout

# Initialize logging
logging.basicConfig(
    level=logging.DEBUG,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    filename='transcript_parser.log'
)

# Model configuration - Only DeepSeek
MODEL_NAME = "deepseek-ai/deepseek-llm-7b"

# Initialize Hugging Face API
if HF_TOKEN:
    try:
        hf_api = HfApi(token=HF_TOKEN)
        HfFolder.save_token(HF_TOKEN)
    except Exception as e:
        logging.error(f"Failed to initialize Hugging Face API: {str(e)}")

# ========== CACHING AND PERFORMANCE OPTIMIZATIONS ==========
executor = ThreadPoolExecutor(max_workers=4)

# Cache model loading
@lru_cache(maxsize=1)
def get_model_and_tokenizer():
    return model_loader.load_model()

# ========== MODEL LOADER ==========
class ModelLoader:
    def __init__(self):
        self.model = None
        self.tokenizer = None
        self.loaded = False
        self.loading = False
        self.error = None
        self.device = "cuda" if torch.cuda.is_available() else "cpu"
    
    def load_model(self, progress: gr.Progress = None) -> Tuple[Optional[AutoModelForCausalLM], Optional[AutoTokenizer]]:
        """Lazy load the model with progress feedback"""
        try:
            if progress:
                progress(0.1, desc="Checking GPU availability...")
            
            # Clear CUDA cache first
            torch.cuda.empty_cache()
            
            if progress:
                progress(0.2, desc="Loading tokenizer...")
                
            tokenizer = AutoTokenizer.from_pretrained(
                MODEL_NAME,
                trust_remote_code=True
            )
            
            if progress:
                progress(0.5, desc="Loading model (this may take a few minutes)...")
                
            # More robust model loading
            model_kwargs = {
                "trust_remote_code": True,
                "torch_dtype": torch.float16 if self.device == "cuda" else torch.float32,
                "device_map": "auto" if self.device == "cuda" else None,
                "low_cpu_mem_usage": True,
                "offload_folder": "offload"  # For handling large models
            }
            
            try:
                model = AutoModelForCausalLM.from_pretrained(
                    MODEL_NAME,
                    **model_kwargs
                )
            except torch.cuda.OutOfMemoryError:
                # Fallback to CPU if GPU OOM
                model_kwargs["device_map"] = None
                model = AutoModelForCausalLM.from_pretrained(
                    MODEL_NAME,
                    **model_kwargs
                ).to('cpu')
                self.device = 'cpu'
                
            # Verify model is responsive
            test_input = tokenizer("Test", return_tensors="pt").to(self.device)
            _ = model.generate(**test_input, max_new_tokens=1)
            
            self.model = model.eval()
            self.tokenizer = tokenizer
            self.loaded = True
            
            return model, tokenizer
            
        except Exception as e:
            self.error = f"Model loading failed: {str(e)}"
            logging.error(self.error)
            return None, None

# Initialize model loader
model_loader = ModelLoader()

# ========== UTILITY FUNCTIONS ==========
def generate_session_token() -> str:
    """Generate a random session token for user identification."""
    alphabet = string.ascii_letters + string.digits
    return ''.join(secrets.choice(alphabet) for _ in range(SESSION_TOKEN_LENGTH))

def sanitize_input(text: str) -> str:
    """Sanitize user input to prevent XSS and injection attacks."""
    if not text:
        return ""
    # Basic HTML escaping and removal of potentially dangerous characters
    text = html.escape(text.strip())
    # Remove any remaining HTML tags
    text = re.sub(r'<[^>]*>', '', text)
    # Remove potentially dangerous characters
    text = re.sub(r'[^\w\s\-.,!?@#\$%^&*()+=]', '', text)
    return text

def validate_name(name: str) -> str:
    """Validate name input."""
    name = name.strip()
    if not name:
        raise ValueError("Name cannot be empty. Please enter your full name.")
    if len(name) > 100:
        raise ValueError("Name is too long (maximum 100 characters).")
    if any(c.isdigit() for c in name):
        raise ValueError("Name cannot contain numbers.")
    return name

def validate_age(age: Union[int, float, str]) -> int:
    """Validate and convert age input."""
    try:
        age_int = int(age)
        if not MIN_AGE <= age_int <= MAX_AGE:
            raise ValueError(f"Age must be between {MIN_AGE} and {MAX_AGE}.")
        return age_int
    except (ValueError, TypeError):
        raise ValueError("Please enter a valid age number.")

def validate_file(file_obj) -> None:
    """Validate uploaded file."""
    if not file_obj:
        raise ValueError("Please upload a file first")
    
    file_ext = os.path.splitext(file_obj.name)[1].lower()
    if file_ext not in ALLOWED_FILE_TYPES:
        raise ValueError(f"Invalid file type. Allowed types: {', '.join(ALLOWED_FILE_TYPES)}")
    
    file_size = os.path.getsize(file_obj.name) / (1024 * 1024)  # MB
    if file_size > MAX_FILE_SIZE_MB:
        raise ValueError(f"File too large. Maximum size is {MAX_FILE_SIZE_MB}MB.")

# ========== TEXT EXTRACTION FUNCTIONS ==========
def extract_text_from_file(file_path: str, file_ext: str) -> str:
    """Enhanced text extraction with better error handling and fallbacks."""
    text = ""
    
    try:
        if file_ext == '.pdf':
            # First try PyMuPDF for text extraction
            try:
                doc = fitz.open(file_path)
                for page in doc:
                    text += page.get_text("text") + '\n'
                if not text.strip():
                    raise ValueError("PyMuPDF returned empty text - the PDF may be image-based")
            except Exception as e:
                logging.warning(f"PyMuPDF failed: {str(e)}. Trying OCR fallback...")
                text = extract_text_from_pdf_with_ocr(file_path)
        
        elif file_ext in ['.png', '.jpg', '.jpeg']:
            text = extract_text_with_ocr(file_path)
            
        # Clean up the extracted text
        text = clean_extracted_text(text)
        
        if not text.strip():
            raise ValueError("No text could be extracted. Please ensure the file is clear and readable.")
            
        return text
    
    except Exception as e:
        logging.error(f"Text extraction error: {str(e)}")
        raise gr.Error(f"Failed to extract text: {str(e)}\n\nTIPS:\n1. For PDFs, try saving as a different PDF format\n2. For images, ensure they are high-quality and well-lit\n3. Try cropping to just the transcript area")

def extract_text_from_pdf_with_ocr(file_path: str) -> str:
    """Fallback PDF text extraction using OCR."""
    text = ""
    try:
        doc = fitz.open(file_path)
        for page in doc:
            pix = page.get_pixmap()
            img = Image.open(io.BytesIO(pix.tobytes()))
            # Preprocess image for better OCR
            img = img.convert('L')  # Grayscale
            img = img.point(lambda x: 0 if x < 128 else 255)  # Binarize
            text += pytesseract.image_to_string(img, config='--psm 6 --oem 3') + '\n'
    except Exception as e:
        raise ValueError(f"PDF OCR failed: {str(e)}. The PDF may be password protected or corrupted.")
    return text

def extract_text_with_ocr(file_path: str) -> str:
    """Extract text from image files using OCR with preprocessing."""
    try:
        image = Image.open(file_path)
        
        # Enhanced preprocessing
        image = image.convert('L')  # Convert to grayscale
        image = image.point(lambda x: 0 if x < 128 else 255, '1')  # Thresholding
        
        # Custom Tesseract configuration
        custom_config = r'--oem 3 --psm 6'
        text = pytesseract.image_to_string(image, config=custom_config)
        return text
    except Exception as e:
        raise ValueError(f"OCR processing failed: {str(e)}. Please ensure the image is clear and not blurry.")

def clean_extracted_text(text: str) -> str:
    """Clean and normalize the extracted text."""
    # Remove multiple spaces and newlines
    text = re.sub(r'\s+', ' ', text).strip()
    
    # Fix common OCR errors
    replacements = {
        '|': 'I',
        '‘': "'",
        '’': "'",
        '“': '"',
        '”': '"',
        'fi': 'fi',
        'fl': 'fl'
    }
    
    for wrong, right in replacements.items():
        text = text.replace(wrong, right)
    
    return text

def remove_sensitive_info(text: str) -> str:
    """Remove potentially sensitive information from transcript text."""
    # Remove social security numbers
    text = re.sub(r'\b\d{3}-\d{2}-\d{4}\b', '[REDACTED]', text)
    # Remove student IDs (assuming 6-9 digit numbers)
    text = re.sub(r'\b\d{6,9}\b', '[ID]', text)
    # Remove email addresses
    text = re.sub(r'\b[A-Za-z0-9._%+-]+@[A-Za-z9.-]+\.[A-Z|a-z]{2,}\b', '[EMAIL]', text)
    return text

def validate_parsed_data(data: Dict) -> bool:
    """Validate the structure of parsed transcript data"""
    required_student_fields = ['name', 'current_grade']
    required_course_fields = ['description', 'credits']
    
    if 'student_info' not in data:
        return False
    if not all(field in data['student_info'] for field in required_student_fields):
        return False
    if 'course_history' not in data or not isinstance(data['course_history'], list):
        return False
    if len(data['course_history']) > 0:
        if not all(field in data['course_history'][0] for field in required_course_fields):
            return False
    return True

# ========== TRANSCRIPT PARSING ==========
class TranscriptParser:
    def __init__(self):
        self.student_data = {}
        self.requirements = {}
        self.current_courses = []
        self.course_history = []
        self.graduation_status = {}
        
    def parse_transcript(self, text: str) -> Dict:
        """Parse Miami-Dade formatted transcripts with updated regex patterns."""
        try:
            # First try structured parsing for Miami-Dade format
            if "Graduation Progress Summary" in text or "Miami-Dade" in text:
                return self._parse_miami_dade_format(text)
            else:
                # Fall back to AI parsing if not Miami-Dade format
                return parse_transcript_with_ai_fallback(text)
                
        except Exception as e:
            logging.error(f"Error parsing transcript: {str(e)}")
            raise ValueError(f"Couldn't parse transcript: {str(e)}")

    def _parse_miami_dade_format(self, text: str, strict_mode: bool = False) -> Dict:
        """Parse Miami-Dade County Public Schools transcripts."""
        # Initialize PDF reader from text (simulating the PDF structure)
        lines = [line.strip() for line in text.split('\n') if line.strip()]
        
        # Initialize data structure
        data = {
            'student_info': {},
            'graduation_requirements': [],
            'course_history': [],
            'summary': {},
            'format': 'miami_dade'  # Add format identifier
        }
        
        # Parse student information with more robust checks
        student_info_lines = []
        for i, line in enumerate(lines):
            logging.debug(f"Processing line: {line}")  # Added debug logging
            if "DORAL ACADEMY HIGH SCHOOL" in line:
                # Get the next 4 lines (or fewer if we're at the end)
                student_info_lines = lines[i:i+5]
                break
        
        if student_info_lines:
            try:
                # Parse school and cohort info - more defensive
                school_info_parts = student_info_lines[0].split('|')
                if len(school_info_parts) > 2:
                    data['student_info']['school'] = school_info_parts[1].strip() if len(school_info_parts) > 1 else ''
                    data['student_info']['district'] = school_info_parts[2].strip() if len(school_info_parts) > 2 else ''
                
                # Parse student name and ID - more defensive
                if len(student_info_lines) > 1:
                    name_id_line = student_info_lines[1].split('-')
                    if len(name_id_line) > 1:
                        name_parts = name_id_line[1].split(',')
                        if len(name_parts) > 1:
                            data['student_info']['student_id'] = name_id_line[0].strip()
                            data['student_info']['student_name'] = name_parts[1].strip() + " " + name_parts[0].strip()
                
                # Parse academic info - more defensive
                if len(student_info_lines) > 2:
                    academic_info = student_info_lines[2].split('|')
                    if len(academic_info) > 5:
                        data['student_info']['current_grade'] = academic_info[1].split(':')[1].strip() if ':' in academic_info[1] else ''
                        data['student_info']['graduation_year'] = academic_info[2].strip()
                        data['student_info']['weighted_gpa'] = academic_info[3].split(':')[1].strip() if ':' in academic_info[3] else ''
                        data['student_info']['community_service_date'] = academic_info[4].split(':')[1].strip() if ':' in academic_info[4] else ''
                        data['student_info']['total_credits_earned'] = academic_info[5].split(':')[1].strip() if ':' in academic_info[5] else ''
                        
                # Validate we got the essential student info
                if not data['student_info'].get('student_name'):
                    logging.warning("Failed to parse student name")
                    if strict_mode:
                        raise ValueError("Could not parse student name from transcript")
                        
            except Exception as e:
                logging.warning(f"Error parsing student info: {str(e)}")
                if strict_mode:
                    raise
        
        # Parse graduation requirements
        requirements_start = None
        requirements_end = None
        for i, line in enumerate(lines):
            logging.debug(f"Processing line: {line}")  # Added debug logging
            if "Code" in line and "Description" in line and "Required" in line:
                requirements_start = i + 1
            if requirements_start and "Total" in line:
                requirements_end = i
                break
        
        if requirements_start and requirements_end:
            for line in lines[requirements_start:requirements_end]:
                try:
                    if '|' in line:
                        parts = [p.strip() for p in line.split('|') if p.strip()]
                        if len(parts) >= 6:
                            req = {
                                'code': parts[0],
                                'description': parts[1],
                                'required': parts[2],
                                'waived': parts[3],
                                'completed': parts[4],
                                'status': parts[5]
                            }
                            data['graduation_requirements'].append(req)
                except Exception as e:
                    logging.warning(f"Error parsing requirement line: {line} - {str(e)}")
                    if strict_mode:
                        raise
            
            # Parse total line
            try:
                total_line = lines[requirements_end]
                total_parts = [p.strip() for p in total_line.split('|') if p.strip()]
                if len(total_parts) >= 5:
                    data['summary']['total_required'] = total_parts[1]
                    data['summary']['total_waived'] = total_parts[2]
                    data['summary']['total_completed'] = total_parts[3]
                    data['summary']['completion_percentage'] = total_parts[4]
            except Exception as e:
                logging.warning(f"Error parsing requirements summary: {str(e)}")
                if strict_mode:
                    raise
        
        # Parse course history
        course_history_start = None
        for i, line in enumerate(lines):
            logging.debug(f"Processing line: {line}")  # Added debug logging
            if "Requirement" in line and "School Year" in line and "GradeLv1" in line:
                course_history_start = i + 1
                break
        
        if course_history_start:
            current_requirement = None
            for line in lines[course_history_start:]:
                try:
                    if '|' in line:
                        parts = [p.strip() for p in line.split('|') if p.strip()]
                        
                        # Check if this is a new requirement line
                        if len(parts) >= 2 and parts[0] and parts[0] in [req['code'] for req in data['graduation_requirements']]:
                            current_requirement = parts[0]
                            parts = parts[1:]  # Remove the requirement code
                        
                        if len(parts) >= 9:
                            course = {
                                'requirement': current_requirement,
                                'school_year': parts[0],
                                'grade_level': parts[1],
                                'course_number': parts[2],
                                'description': parts[3],
                                'term': parts[4],
                                'district_number': parts[5],
                                'fg': parts[6],
                                'included': parts[7],
                                'credits': parts[8]
                            }
                            data['course_history'].append(course)
                except Exception as e:
                    logging.warning(f"Error parsing course line: {line} - {str(e)}")
                    if strict_mode:
                        raise
        
        # Calculate graduation status
        try:
            if data['summary'].get('total_required') and data['summary'].get('total_completed'):
                graduation_status = {
                    'total_required_credits': float(data['summary']['total_required']),
                    'total_completed_credits': float(data['summary']['total_completed']),
                    'percent_complete': float(data['summary']['completion_percentage'].replace('%', '')),
                    'remaining_credits': float(data['summary']['total_required']) - float(data['summary']['total_completed']),
                    'on_track': float(data['summary']['completion_percentage'].replace('%', '')) >= 75.0
                }
                data['graduation_status'] = graduation_status
        except Exception as e:
            logging.warning(f"Error calculating graduation status: {str(e)}")
            if strict_mode:
                raise
        
        return data

def format_transcript_output(data: Dict) -> str:
    """Enhanced formatting for transcript output with format awareness"""
    output = []
    
    # Student Info Section
    student = data.get("student_info", {})
    output.append(f"## Student Transcript Summary\n{'='*50}")
    output.append(f"**Name:** {student.get('name', 'Unknown')}")
    output.append(f"**Student ID:** {student.get('id', 'Unknown')}")
    output.append(f"**Current Grade:** {student.get('current_grade', 'Unknown')}")
    output.append(f"**Graduation Year:** {student.get('graduation_year', 'Unknown')}")
    
    if 'unweighted_gpa' in student and 'weighted_gpa' in student:
        output.append(f"**Unweighted GPA:** {student['unweighted_gpa']}")
        output.append(f"**Weighted GPA:** {student['weighted_gpa']}")
    elif 'gpa' in student:
        output.append(f"**GPA:** {student['gpa']}")
    
    if 'total_credits' in student:
        output.append(f"**Total Credits Earned:** {student['total_credits']}")
    if 'community_service_hours' in student:
        output.append(f"**Community Service Hours:** {student['community_service_hours']}")
    
    output.append("")
    
    # Graduation Requirements Section (for Miami-Dade format)
    if data.get('format') == 'miami_dade':
        grad_status = data.get("graduation_status", {})
        output.append(f"## Graduation Progress\n{'='*50}")
        output.append(f"**Overall Completion:** {grad_status.get('percent_complete', 0)}%")
        output.append(f"**Credits Required:** {grad_status.get('total_required_credits', 0)}")
        output.append(f"**Credits Completed:** {grad_status.get('total_completed_credits', 0)}")
        output.append(f"**Credits Remaining:** {grad_status.get('remaining_credits', 0)}")
        output.append(f"**On Track to Graduate:** {'Yes' if grad_status.get('on_track', False) else 'No'}\n")
        
        # Detailed Requirements
        output.append("### Detailed Requirements:")
        for req in data.get("graduation_requirements", []):
            output.append(
                f"- **{req['code']}**: {req['description']}\n"
                f"  Required: {req['required']} | Completed: {req['completed']} | "
                f"Status: {req['status']}"
            )
        output.append("")
    
    # Current Courses
    if any(c.get('credits', '') == 'inProgress' for c in data.get("course_history", [])):
        output.append("## Current Courses (In Progress)\n" + '='*50)
        for course in data["course_history"]:
            if course.get('credits', '') == 'inProgress':
                output.append(
                    f"- **{course['course_number']} {course['description']}**\n"
                    f"  Category: {course['requirement']} | "
                    f"Grade Level: {course['grade_level']} | "
                    f"Term: {course['term']} | Credits: {course['credits']}"
                )
        output.append("")
    
    # Course History by Year
    courses_by_year = defaultdict(list)
    for course in data.get("course_history", []):
        if course.get("school_year"):
            courses_by_year[course["school_year"]].append(course)
    
    if courses_by_year:
        output.append("## Course History\n" + '='*50)
        for year in sorted(courses_by_year.keys()):
            output.append(f"\n### {year}")
            for course in courses_by_year[year]:
                output.append(
                    f"- **{course.get('course_number', '')} {course.get('description', 'Unnamed course')}**\n"
                    f"  Subject: {course.get('requirement', 'N/A')} | "
                    f"Grade: {course.get('fg', 'N/A')} | "
                    f"Credits: {course.get('credits', 'N/A')}"
                )
    
    return '\n'.join(output)

def parse_transcript_with_ai_fallback(text: str, progress=gr.Progress()) -> Dict:
    """More robust AI parsing with better error handling"""
    try:
        text = remove_sensitive_info(text[:20000])  # Increased limit
        
        # Improved prompt with examples
        prompt = f"""Extract academic transcript data as JSON. Follow this structure:
        
        Example Input:
        Student ID: 1234567 Name: DOE, JOHN Current Grade: 12 YOG: 2024
        Unweighted GPA: 3.5 Weighted GPA: 4.2 Total Credits: 24.5
        
        Example Output:
        {{
            "student_info": {{
                "name": "John Doe",
                "id": "1234567",
                "current_grade": "12",
                "graduation_year": "2024",
                "unweighted_gpa": 3.5,
                "weighted_gpa": 4.2,
                "total_credits": 24.5
            }},
            "course_history": [
                {{
                    "course_code": "MATH101",
                    "description": "Algebra I",
                    "grade": "A",
                    "credits": 1.0,
                    "school_year": "2022-2023"
                }}
            ]
        }}
        
        Actual Transcript:
        {text}
        """
        
        if progress:
            progress(0.3, desc="Processing with AI...")
        
        model, tokenizer = get_model_and_tokenizer()
        if model is None:
            raise ValueError("Model not loaded")
            
        inputs = tokenizer(prompt, return_tensors="pt", truncation=True).to(model_loader.device)
        
        outputs = model.generate(
            **inputs,
            max_new_tokens=2500,
            temperature=0.3,  # Lower for more consistent results
            do_sample=True,
            top_p=0.9,
            repetition_penalty=1.2
        )
        
        response = tokenizer.decode(outputs[0], skip_special_tokens=True)
        
        # More robust JSON extraction
        try:
            if '```json' in response:
                json_str = response.split('```json')[1].split('```')[0].strip()
            else:
                json_str = response.split('{', 1)[1].rsplit('}', 1)[0]
                json_str = '{' + json_str + '}'
                
            parsed_data = json.loads(json_str)
            
            # Validate required fields
            if not all(k in parsed_data for k in ["student_info", "course_history"]):
                raise ValueError("Missing required fields in AI response")
                
            return parsed_data
            
        except Exception as e:
            logging.error(f"JSON parsing failed: {str(e)}")
            raise ValueError(f"AI returned invalid format. Please try again.")
            
    except Exception as e:
        logging.error(f"AI parsing error: {str(e)}")
        raise gr.Error(f"Failed to parse transcript: {str(e)}")

def parse_transcript_with_ai(text: str, progress=gr.Progress()) -> Dict:
    """Enhanced AI parsing with fallback to structured parsing"""
    try:
        # First try structured parsing
        if progress:
            progress(0.1, desc="Attempting structured parsing...")
        
        parser = TranscriptParser()
        parsed_data = parser.parse_transcript(text)
        
        # Validate the parsed data
        if not validate_parsed_data(parsed_data):
            raise ValueError("Structured parsing returned incomplete data")
            
        if progress:
            progress(0.8, desc="Formatting results...")
        
        return parsed_data
        
    except Exception as e:
        logging.warning(f"Structured parsing failed, falling back to AI: {str(e)}")
        
        # Fall back to AI parsing if structured parsing fails
        return parse_transcript_with_ai_fallback(text, progress)

async def parse_transcript_async(file_obj, progress=gr.Progress()) -> Tuple[str, Optional[Dict]]:
    """Async wrapper for transcript parsing"""
    loop = asyncio.get_event_loop()
    return await loop.run_in_executor(executor, parse_transcript, file_obj, progress)

def parse_transcript(file_obj, progress=gr.Progress()) -> Tuple[str, Optional[Dict]]:
    """Main function to parse transcript files with better error handling"""
    try:
        if not file_obj:
            raise ValueError("Please upload a file first")
            
        validate_file(file_obj)
        file_ext = os.path.splitext(file_obj.name)[1].lower()
        
        # Extract text from file with better error reporting
        if progress:
            progress(0.2, desc="Extracting text from file...")
        
        text = extract_text_from_file(file_obj.name, file_ext)
        
        if not text.strip():
            raise ValueError("No text could be extracted from the file. The file may be corrupted or in an unsupported format.")
        
        # Try structured parsing first
        if progress:
            progress(0.4, desc="Attempting structured parsing...")
            
        parser = TranscriptParser()
        try:
            parsed_data = parser.parse_transcript(text)
            if validate_parsed_data(parsed_data):
                if progress:
                    progress(0.9, desc="Formatting results...")
                return format_transcript_output(parsed_data), parsed_data
        except Exception as e:
            logging.warning(f"Structured parsing failed: {str(e)}")
        
        # Fall back to AI if structured fails
        if progress:
            progress(0.5, desc="Using AI analysis...")
            
        parsed_data = parse_transcript_with_ai_fallback(text, progress)
        return format_transcript_output(parsed_data), parsed_data
        
    except Exception as e:
        error_msg = f"Error processing transcript: {str(e)}"
        # Add specific troubleshooting tips
        if "PDF" in str(e):
            error_msg += "\n\nTIPS:\n1. Try converting to image (screenshot)\n2. Ensure text is selectable in PDF\n3. Try a different PDF reader"
        elif "image" in str(e).lower():
            error_msg += "\n\nTIPS:\n1. Use high contrast images\n2. Crop to just the transcript\n3. Ensure good lighting"
        elif "AI" in str(e):
            error_msg += "\n\nTIPS:\n1. Try a smaller section of the transcript\n2. Check for sensitive info that may be redacted\n3. Try again later"
            
        logging.error(error_msg)
        return error_msg, None

# ========== LEARNING STYLE QUIZ ==========
class LearningStyleQuiz:
    def __init__(self):
        self.questions = [
            "When you study for a test, you prefer to:",
            "When you need directions to a new place, you prefer:",
            "When you learn a new skill, you prefer to:",
            "When you're trying to concentrate, you:",
            "When you meet new people, you remember them by:",
            "When you're assembling furniture or a gadget, you:",
            "When choosing a restaurant, you rely most on:",
            "When you're in a waiting room, you typically:",
            "When giving someone instructions, you tend to:",
            "When you're trying to recall information, you:",
            "When you're at a museum or exhibit, you:",
            "When you're learning a new language, you prefer:",
            "When you're taking notes in class, you:",
            "When you're explaining something complex, you:",
            "When you're at a party, you enjoy:",
            "When you're trying to remember a phone number, you:",
            "When you're relaxing, you prefer to:",
            "When you're learning to use new software, you:",
            "When you're giving a presentation, you rely on:",
            "When you're solving a difficult problem, you:"
        ]
        
        self.options = [
            ["Read the textbook (Reading/Writing)", "Listen to lectures (Auditory)", "Use diagrams/charts (Visual)", "Practice problems (Kinesthetic)"],
            ["Look at a map (Visual)", "Have someone tell you (Auditory)", "Write down directions (Reading/Writing)", "Try walking/driving there (Kinesthetic)"],
            ["Read instructions (Reading/Writing)", "Have someone show you (Visual)", "Listen to explanations (Auditory)", "Try it yourself (Kinesthetic)"],
            ["Need quiet (Reading/Writing)", "Need background noise (Auditory)", "Need to move around (Kinesthetic)", "Need visual stimulation (Visual)"],
            ["Their face (Visual)", "Their name (Auditory)", "What you talked about (Reading/Writing)", "What you did together (Kinesthetic)"],
            ["Read the instructions carefully (Reading/Writing)", "Look at the diagrams (Visual)", "Ask someone to explain (Auditory)", "Start putting pieces together (Kinesthetic)"],
            ["Online photos of the food (Visual)", "Recommendations from friends (Auditory)", "Reading the menu online (Reading/Writing)", "Remembering how it felt to eat there (Kinesthetic)"],
            ["Read magazines (Reading/Writing)", "Listen to music (Auditory)", "Watch TV (Visual)", "Fidget or move around (Kinesthetic)"],
            ["Write them down (Reading/Writing)", "Explain verbally (Auditory)", "Demonstrate (Visual)", "Guide them physically (Kinesthetic)"],
            ["See written words in your mind (Visual)", "Hear the information in your head (Auditory)", "Write it down to remember (Reading/Writing)", "Associate it with physical actions (Kinesthetic)"],
            ["Read all the descriptions (Reading/Writing)", "Listen to audio guides (Auditory)", "Look at the displays (Visual)", "Touch interactive exhibits (Kinesthetic)"],
            ["Study grammar rules (Reading/Writing)", "Listen to native speakers (Auditory)", "Use flashcards with images (Visual)", "Practice conversations (Kinesthetic)"],
            ["Write detailed paragraphs (Reading/Writing)", "Record the lecture (Auditory)", "Draw diagrams and charts (Visual)", "Doodle while listening (Kinesthetic)"],
            ["Write detailed steps (Reading/Writing)", "Explain verbally with examples (Auditory)", "Draw diagrams (Visual)", "Use physical objects to demonstrate (Kinesthetic)"],
            ["Conversations with people (Auditory)", "Watching others or the environment (Visual)", "Writing notes or texting (Reading/Writing)", "Dancing or physical activities (Kinesthetic)"],
            ["See the numbers in your head (Visual)", "Say them aloud (Auditory)", "Write them down (Reading/Writing)", "Dial them on a keypad (Kinesthetic)"],
            ["Read a book (Reading/Writing)", "Listen to music (Auditory)", "Watch TV/movies (Visual)", "Do something physical (Kinesthetic)"],
            ["Read the manual (Reading/Writing)", "Ask someone to show you (Visual)", "Call tech support (Auditory)", "Experiment with the software (Kinesthetic)"],
            ["Detailed notes (Reading/Writing)", "Verbal explanations (Auditory)", "Visual slides (Visual)", "Physical demonstrations (Kinesthetic)"],
            ["Write out possible solutions (Reading/Writing)", "Talk through it with someone (Auditory)", "Draw diagrams (Visual)", "Build a model or prototype (Kinesthetic)"]
        ]
        
        self.learning_styles = {
            "Visual": {
                "description": "Visual learners prefer using images, diagrams, and spatial understanding.",
                "tips": [
                    "Use color coding in your notes",
                    "Create mind maps and diagrams",
                    "Watch educational videos",
                    "Use flashcards with images",
                    "Highlight important information in different colors"
                ],
                "careers": [
                    "Graphic Designer", "Architect", "Photographer", 
                    "Engineer", "Surgeon", "Pilot"
                ]
            },
            "Auditory": {
                "description": "Auditory learners learn best through listening and speaking.",
                "tips": [
                    "Record lectures and listen to them",
                    "Participate in study groups",
                    "Explain concepts out loud to yourself",
                    "Use rhymes or songs to remember information",
                    "Listen to educational podcasts"
                ],
                "careers": [
                    "Musician", "Journalist", "Lawyer",
                    "Psychologist", "Teacher", "Customer Service"
                ]
            },
            "Reading/Writing": {
                "description": "These learners prefer information displayed as words.",
                "tips": [
                    "Write detailed notes",
                    "Create summaries in your own words",
                    "Read textbooks and articles",
                    "Make lists to organize information",
                    "Rewrite your notes to reinforce learning"
                ],
                "careers": [
                    "Writer", "Researcher", "Editor",
                    "Accountant", "Programmer", "Historian"
                ]
            },
            "Kinesthetic": {
                "description": "Kinesthetic learners learn through movement and hands-on activities.",
                "tips": [
                    "Use hands-on activities",
                    "Take frequent movement breaks",
                    "Create physical models",
                    "Associate information with physical actions",
                    "Study while walking or pacing"
                ],
                "careers": [
                    "Athlete", "Chef", "Mechanic",
                    "Dancer", "Physical Therapist", "Carpenter"
                ]
            }
        }
    
    def evaluate_quiz(self, *answers) -> str:
        """Evaluate quiz answers and generate enhanced results."""
        answers = list(answers)  # Convert tuple to list
        if len(answers) != len(self.questions):
            raise gr.Error("Please answer all questions before submitting")
        
        scores = {style: 0 for style in self.learning_styles}
        
        for i, answer in enumerate(answers):
            if not answer:
                continue  # Skip unanswered questions
                
            for j, style in enumerate(self.learning_styles):
                if answer == self.options[i][j]:
                    scores[style] += 1
                    break
        
        total_answered = sum(1 for ans in answers if ans)
        if total_answered == 0:
            raise gr.Error("No answers provided")
        
        percentages = {style: (score/total_answered)*100 for style, score in scores.items()}
        sorted_styles = sorted(scores.items(), key=lambda x: x[1], reverse=True)
        
        # Generate enhanced results report
        result = "## Your Learning Style Results\n\n"
        result += "### Scores:\n"
        for style, score in sorted_styles:
            result += f"- **{style}**: {score}/{total_answered} ({percentages[style]:.1f}%)\n"
        
        max_score = max(scores.values())
        primary_styles = [style for style, score in scores.items() if score == max_score]
        
        result += "\n### Analysis:\n"
        if len(primary_styles) == 1:
            primary_style = primary_styles[0]
            style_info = self.learning_styles[primary_style]
            
            result += f"Your primary learning style is **{primary_style}**\n\n"
            result += f"**{primary_style} Characteristics**:\n"
            result += f"{style_info['description']}\n\n"
            
            result += "**Recommended Study Strategies**:\n"
            for tip in style_info['tips']:
                result += f"- {tip}\n"
            
            result += "\n**Potential Career Paths**:\n"
            for career in style_info['careers'][:6]:
                result += f"- {career}\n"
            
            # Add complementary strategies
            complementary = [s for s in sorted_styles if s[0] != primary_style][0][0]
            result += f"\nYou might also benefit from some **{complementary}** strategies:\n"
            for tip in self.learning_styles[complementary]['tips'][:3]:
                result += f"- {tip}\n"
        else:
            result += "You have multiple strong learning styles:\n"
            for style in primary_styles:
                result += f"- **{style}**\n"
            
            result += "\n**Combined Learning Strategies**:\n"
            result += "You may benefit from combining different learning approaches:\n"
            for style in primary_styles:
                result += f"\n**{style}** techniques:\n"
                for tip in self.learning_styles[style]['tips'][:2]:
                    result += f"- {tip}\n"
                
                result += f"\n**{style}** career suggestions:\n"
                for career in self.learning_styles[style]['careers'][:3]:
                    result += f"- {career}\n"
        
        return result

# Initialize quiz instance
learning_style_quiz = LearningStyleQuiz()

# ========== PROFILE MANAGEMENT ==========
class ProfileManager:
    def __init__(self):
        self.profiles_dir = Path(PROFILES_DIR)
        self.profiles_dir.mkdir(exist_ok=True, parents=True)
        self.current_session = None
    
    def set_session(self, session_token: str) -> None:
        """Set the current session token."""
        self.current_session = session_token
    
    def get_profile_path(self, name: str) -> Path:
        """Get profile path with session token if available."""
        if self.current_session:
            # Hash the name for security
            name_hash = hashlib.sha256(name.encode()).hexdigest()[:16]
            return self.profiles_dir / f"{name_hash}_{self.current_session}_profile.json"
        return self.profiles_dir / f"{name.replace(' ', '_')}_profile.json"
    
    def save_profile(self, name: str, age: Union[int, str], interests: str, 
                    transcript: Dict, learning_style: str, 
                    movie: str, movie_reason: str, show: str, show_reason: str, 
                    book: str, book_reason: str, character: str, character_reason: str, 
                    blog: str) -> str:
        """Save student profile with better validation messages"""
        try:
            # Validate required fields with specific messages
            if not name.strip():
                raise ValueError("Name cannot be empty. Please enter your full name.")
            if len(name) > 100:
                raise ValueError("Name is too long (maximum 100 characters).")
            if any(c.isdigit() for c in name):
                raise ValueError("Name cannot contain numbers.")
            
            try:
                age_int = int(age)
                if not MIN_AGE <= age_int <= MAX_AGE:
                    raise ValueError(f"Age must be between {MIN_AGE} and {MAX_AGE}.")
            except (ValueError, TypeError):
                raise ValueError("Please enter a valid age number.")
            
            if not interests.strip():
                raise ValueError("Please describe at least one interest or hobby.")
            
            if not transcript:
                raise ValueError("Please complete the transcript analysis first.")
            
            # Validate learning style quiz completion
            if not learning_style or "Your primary learning style is:" not in learning_style:
                raise ValueError("Please complete the learning style quiz first.")
            
            # Prepare favorites data
            favorites = {
                "movie": sanitize_input(movie),
                "movie_reason": sanitize_input(movie_reason),
                "show": sanitize_input(show),
                "show_reason": sanitize_input(show_reason),
                "book": sanitize_input(book),
                "book_reason": sanitize_input(book_reason),
                "character": sanitize_input(character),
                "character_reason": sanitize_input(character_reason)
            }
            
            # Prepare full profile data
            data = {
                "name": name,
                "age": age_int,
                "interests": sanitize_input(interests),
                "transcript": transcript if transcript else {},
                "learning_style": learning_style if learning_style else "Not assessed",
                "favorites": favorites,
                "blog": sanitize_input(blog) if blog else "",
                "session_token": self.current_session,
                "last_updated": time.time()
            }
            
            # Save to JSON file
            filepath = self.get_profile_path(name)
            
            with open(filepath, "w", encoding='utf-8') as f:
                json.dump(data, f, indent=2, ensure_ascii=False)
            
            # Upload to HF Hub if token is available
            if HF_TOKEN and 'hf_api' in globals():
                try:
                    hf_api.upload_file(
                        path_or_fileobj=filepath,
                        path_in_repo=f"profiles/{filepath.name}",
                        repo_id="your-username/student-learning-assistant",
                        repo_type="dataset"
                    )
                except Exception as e:
                    logging.error(f"Failed to upload to HF Hub: {str(e)}")
            
            return self._generate_profile_summary(data)
        
        except Exception as e:
            logging.error(f"Profile validation error: {str(e)}")
            raise gr.Error(f"Couldn't save profile: {str(e)}")
    
    def load_profile(self, name: str = None, session_token: str = None) -> Dict:
        """Load profile by name or return the first one found."""
        try:
            if session_token:
                profile_pattern = f"*{session_token}_profile.json"
            else:
                profile_pattern = "*.json"
            
            profiles = list(self.profiles_dir.glob(profile_pattern))
            if not profiles:
                return {}
            
            if name:
                # Find profile by name (hashed)
                name_hash = hashlib.sha256(name.encode()).hexdigest()[:16]
                if session_token:
                    profile_file = self.profiles_dir / f"{name_hash}_{session_token}_profile.json"
                else:
                    profile_file = self.profiles_dir / f"{name_hash}_profile.json"
                
                if not profile_file.exists():
                    # Try loading from HF Hub
                    if HF_TOKEN and 'hf_api' in globals():
                        try:
                            hf_api.download_file(
                                path_in_repo=f"profiles/{profile_file.name}",
                                repo_id="your-username/student-learning-assistant",
                                repo_type="dataset",
                                local_dir=self.profiles_dir
                            )
                        except:
                            raise gr.Error(f"No profile found for {name}")
                    else:
                        raise gr.Error(f"No profile found for {name}")
            else:
                # Load the first profile found
                profile_file = profiles[0]
            
            with open(profile_file, "r", encoding='utf-8') as f:
                profile_data = json.load(f)
                # Check session timeout
                if time.time() - profile_data.get('last_updated', 0) > SESSION_TIMEOUT:
                    raise gr.Error("Session expired. Please start a new session.")
                return profile_data
        
        except Exception as e:
            logging.error(f"Error loading profile: {str(e)}")
            return {}
    
    def list_profiles(self, session_token: str = None) -> List[str]:
        """List all available profile names for the current session."""
        if session_token:
            profiles = list(self.profiles_dir.glob(f"*{session_token}_profile.json"))
        else:
            profiles = list(self.profiles_dir.glob("*.json"))
        
        # Extract just the name part (without session token)
        profile_names = []
        for p in profiles:
            with open(p, "r", encoding='utf-8') as f:
                try:
                    data = json.load(f)
                    profile_names.append(data.get('name', p.stem))
                except json.JSONDecodeError:
                    continue
        
        return profile_names
    
    def _generate_profile_summary(self, data: Dict) -> str:
        """Generate markdown summary of the profile."""
        transcript = data.get("transcript", {})
        favorites = data.get("favorites", {})
        
        # Extract just the learning style name
        learning_style = data.get("learning_style", "")
        if "Your primary learning style is:" in learning_style:
            style_match = re.search(r"Your primary learning style is: \*\*(.*?)\*\*", learning_style)
            if style_match:
                learning_style = style_match.group(1)
        
        markdown = f"""## Student Profile: {data['name']}
### Basic Information
- **Age:** {data['age']}
- **Interests:** {data.get('interests', 'Not specified')}
- **Learning Style:** {learning_style}
### Academic Information
{self._format_transcript(transcript)}
### Favorites
- **Movie:** {favorites.get('movie', 'Not specified')}  
  *Reason:* {favorites.get('movie_reason', 'Not specified')}
- **TV Show:** {favorites.get('show', 'Not specified')}  
  *Reason:* {favorites.get('show_reason', 'Not specified')}
- **Book:** {favorites.get('book', 'Not specified')}  
  *Reason:* {favorites.get('book_reason', 'Not specified')}
- **Character:** {favorites.get('character', 'Not specified')}  
  *Reason:* {favorites.get('character_reason', 'Not specified')}
### Personal Blog
{data.get('blog', '_No blog provided_')}
"""
        return markdown
    
    def _format_transcript(self, transcript: Dict) -> str:
        """Format transcript data for display."""
        if not transcript or "course_history" not in transcript:
            return "_No transcript information available_"
        
        display = "#### Course History\n"
        courses_by_year = defaultdict(list)
        for course in transcript.get("course_history", []):
            if course.get("school_year"):
                courses_by_year[course["school_year"]].append(course)
        
        if courses_by_year:
            for year in sorted(courses_by_year.keys()):
                display += f"\n**{year}**\n"
                for course in courses_by_year[year]:
                    display += f"- {course.get('course_code', '')} {course.get('description', 'Unnamed course')}"
                    if 'grade' in course and course['grade']:
                        display += f" (Grade: {course['grade']})"
                    if 'credits' in course:
                        display += f" | Credits: {course['credits']}"
                    display += f" | Category: {course.get('requirement_category', 'N/A')}\n"
        
        if 'student_info' in transcript:
            student = transcript['student_info']
            display += "\n**Academic Summary**\n"
            display += f"- Unweighted GPA: {student.get('unweighted_gpa', 'N/A')}\n"
            display += f"- Weighted GPA: {student.get('weighted_gpa', 'N/A')}\n"
            display += f"- Total Credits: {student.get('total_credits', 'N/A')}\n"
        
        if 'graduation_status' in transcript:
            status = transcript['graduation_status']
            display += "\n**Graduation Progress**\n"
            display += f"- Completion: {status.get('percent_complete', 0)}%\n"
            display += f"- Credits Required: {status.get('total_required_credits', 0)}\n"
            display += f"- Credits Completed: {status.get('total_completed_credits', 0)}\n"
            display += f"- On Track: {'Yes' if status.get('on_track', False) else 'No'}\n"
        
        return display

# Initialize profile manager
profile_manager = ProfileManager()

# ========== AI TEACHING ASSISTANT ==========
class TeachingAssistant:
    def __init__(self):
        self.context_history = []
        self.max_context_length = 5  # Keep last 5 exchanges for context
    
    async def generate_response(self, message: str, history: List[List[Union[str, None]]], session_token: str) -> str:
        """Generate personalized response based on student profile and context."""
        try:
            # Load profile with session token
            profile = profile_manager.load_profile(session_token=session_token)
            if not profile:
                return "Please complete and save your profile first using the previous tabs."
            
            # Update context history
            self._update_context(message, history)
            
            # Extract profile information
            name = profile.get("name", "there")
            learning_style = profile.get("learning_style", "")
            grade_level = profile.get("transcript", {}).get("student_info", {}).get("current_grade", "unknown")
            gpa = profile.get("transcript", {}).get("student_info", {})
            interests = profile.get("interests", "")
            courses = profile.get("transcript", {}).get("course_history", [])
            favorites = profile.get("favorites", {})
            
            # Process message with context
            response = await self._process_message(message, profile)
            
            # Add follow-up suggestions
            if "study" in message.lower() or "learn" in message.lower():
                response += "\n\nWould you like me to suggest a study schedule based on your courses?"
            elif "course" in message.lower() or "class" in message.lower():
                response += "\n\nWould you like help finding resources for any of these courses?"
            
            return response
        
        except Exception as e:
            logging.error(f"Error generating response: {str(e)}")
            return "I encountered an error processing your request. Please try again."
    
    def _update_context(self, message: str, history: List[List[Union[str, None]]]) -> None:
        """Maintain conversation context."""
        self.context_history.append({"role": "user", "content": message})
        if history:
            for h in history[-self.max_context_length:]:
                if h[0]:  # User message
                    self.context_history.append({"role": "user", "content": h[0]})
                if h[1]:  # Assistant message
                    self.context_history.append({"role": "assistant", "content": h[1]})
        
        # Trim to maintain max context length
        self.context_history = self.context_history[-(self.max_context_length*2):]
    
    async def _process_message(self, message: str, profile: Dict) -> str:
        """Process user message with profile context."""
        message_lower = message.lower()
        
        # Greetings
        if any(greet in message_lower for greet in ["hi", "hello", "hey", "greetings"]):
            return f"Hello {profile.get('name', 'there')}! How can I help you with your learning today?"
        
        # Study help
        study_words = ["study", "learn", "prepare", "exam", "test", "homework"]
        if any(word in message_lower for word in study_words):
            return self._generate_study_advice(profile)
        
        # Grade help
        grade_words = ["grade", "gpa", "score", "marks", "results"]
        if any(word in message_lower for word in grade_words):
            return self._generate_grade_advice(profile)
        
        # Interest help
        interest_words = ["interest", "hobby", "passion", "extracurricular"]
        if any(word in message_lower for word in interest_words):
            return self._generate_interest_advice(profile)
        
        # Course help
        course_words = ["courses", "classes", "transcript", "schedule", "subject"]
        if any(word in message_lower for word in course_words):
            return self._generate_course_advice(profile)
        
        # Favorites
        favorite_words = ["movie", "show", "book", "character", "favorite"]
        if any(word in message_lower for word in favorite_words):
            return self._generate_favorites_response(profile)
        
        # General help
        if "help" in message_lower:
            return self._generate_help_response()
        
        # Default response
        return ("I'm your personalized teaching assistant. I can help with study tips, "
                "grade information, course advice, and more. Try asking about how to "
                "study effectively or about your course history.")
    
    def _generate_study_advice(self, profile: Dict) -> str:
        """Generate study advice based on learning style."""
        learning_style = profile.get("learning_style", "")
        response = ""
        
        if "Visual" in learning_style:
            response = ("Based on your visual learning style, I recommend:\n"
                       "- Creating colorful mind maps or diagrams\n"
                       "- Using highlighters to color-code your notes\n"
                       "- Watching educational videos on the topics\n"
                       "- Creating flashcards with images\n\n")
        elif "Auditory" in learning_style:
            response = ("Based on your auditory learning style, I recommend:\n"
                       "- Recording your notes and listening to them\n"
                       "- Participating in study groups to discuss concepts\n"
                       "- Explaining the material out loud to yourself\n"
                       "- Finding podcasts or audio lectures on the topics\n\n")
        elif "Reading/Writing" in learning_style:
            response = ("Based on your reading/writing learning style, I recommend:\n"
                       "- Writing detailed summaries in your own words\n"
                       "- Creating organized outlines of the material\n"
                       "- Reading additional textbooks or articles\n"
                       "- Rewriting your notes to reinforce learning\n\n")
        elif "Kinesthetic" in learning_style:
            response = ("Based on your kinesthetic learning style, I recommend:\n"
                       "- Creating physical models or demonstrations\n"
                       "- Using hands-on activities to learn concepts\n"
                       "- Taking frequent movement breaks while studying\n"
                       "- Associating information with physical actions\n\n")
        else:
            response = ("Here are some general study tips:\n"
                       "- Use the Pomodoro technique (25 min study, 5 min break)\n"
                       "- Space out your study sessions over time\n"
                       "- Test yourself with practice questions\n"
                       "- Teach the material to someone else\n\n")
        
        # Add time management advice
        response += ("**Time Management Tips**:\n"
                    "- Create a study schedule and stick to it\n"
                    "- Prioritize difficult subjects when you're most alert\n"
                    "- Break large tasks into smaller, manageable chunks\n"
                    "- Set specific goals for each study session")
        
        return response
    
    def _generate_grade_advice(self, profile: Dict) -> str:
        """Generate response about grades and GPA."""
        gpa = profile.get("transcript", {}).get("student_info", {})
        courses = profile.get("transcript", {}).get("course_history", [])
        
        response = (f"Your GPA information:\n"
                   f"- Unweighted: {gpa.get('unweighted_gpa', 'N/A')}\n"
                   f"- Weighted: {gpa.get('weighted_gpa', 'N/A')}\n\n")
        
        # Identify any failing grades
        weak_subjects = []
        for course in courses:
            if course.get('grade', '').upper() in ['D', 'F']:
                weak_subjects.append(f"{course.get('course_code', '')} {course.get('description', 'Unknown course')}")
        
        if weak_subjects:
            response += ("**Areas for Improvement**:\n"
                        f"You might want to focus on these subjects: {', '.join(weak_subjects)}\n\n")
        
        response += ("**Grade Improvement Strategies**:\n"
                    "- Meet with your teachers to discuss your performance\n"
                    "- Identify specific areas where you lost points\n"
                    "- Create a targeted study plan for weak areas\n"
                    "- Practice with past exams or sample questions")
        
        return response
    
    def _generate_interest_advice(self, profile: Dict) -> str:
        """Generate response based on student interests."""
        interests = profile.get("interests", "")
        response = f"I see you're interested in: {interests}\n\n"
        
        response += ("**Suggestions**:\n"
                    "- Look for clubs or extracurricular activities related to these interests\n"
                    "- Explore career paths that align with these interests\n"
                    "- Find online communities or forums about these topics\n"
                    "- Consider projects or independent study in these areas")
        
        return response
    
    def _generate_course_advice(self, profile: Dict) -> str:
        """Generate response about courses."""
        courses = profile.get("transcript", {}).get("course_history", [])
        grade_level = profile.get("transcript", {}).get("student_info", {}).get("current_grade", "unknown")
        
        response = "Here's a summary of your courses by year:\n"
        courses_by_year = defaultdict(list)
        for course in courses:
            if course.get("school_year"):
                courses_by_year[course["school_year"]].append(course)
        
        for year in sorted(courses_by_year.keys()):
            response += f"\n**{year}**:\n"
            for course in courses_by_year[year]:
                response += f"- {course.get('course_code', '')} {course.get('description', 'Unnamed course')}"
                if 'grade' in course:
                    response += f" (Grade: {course['grade']})"
                response += "\n"
        
        response += f"\nAs a grade {grade_level} student, you might want to:\n"
        if grade_level in ["9", "10"]:
            response += ("- Focus on building strong foundational skills\n"
                        "- Explore different subjects to find your interests\n"
                        "- Start thinking about college/career requirements")
        elif grade_level in ["11", "12"]:
            response += ("- Focus on courses relevant to your college/career goals\n"
                        "- Consider taking AP or advanced courses if available\n"
                        "- Ensure you're meeting graduation requirements")
        
        return response
    
    def _generate_favorites_response(self, profile: Dict) -> str:
        """Generate response about favorite items."""
        favorites = profile.get("favorites", {})
        response = "I see you enjoy:\n"
        
        if favorites.get('movie'):
            response += f"- Movie: {favorites['movie']} ({favorites.get('movie_reason', 'no reason provided')})\n"
        if favorites.get('show'):
            response += f"- TV Show: {favorites['show']} ({favorites.get('show_reason', 'no reason provided')})\n"
        if favorites.get('book'):
            response += f"- Book: {favorites['book']} ({favorites.get('book_reason', 'no reason provided')})\n"
        if favorites.get('character'):
            response += f"- Character: {favorites['character']} ({favorites.get('character_reason', 'no reason provided')})\n"
        
        response += "\nThese preferences suggest you might enjoy:\n"
        response += "- Similar books/movies in the same genre\n"
        response += "- Creative projects related to these stories\n"
        response += "- Analyzing themes or characters in your schoolwork"
        
        return response
    
    def _generate_help_response(self) -> str:
        """Generate help response with available commands."""
        return ("""I can help with:
- **Study tips**: "How should I study for math?"
- **Grade information**: "What's my GPA?"
- **Course advice**: "Show me my course history"
- **Interest suggestions**: "What clubs match my interests?"
- **General advice**: "How can I improve my grades?"
Try asking about any of these topics!""")

# Initialize teaching assistant
teaching_assistant = TeachingAssistant()

# ========== GRADIO INTERFACE ==========
def create_interface():
    with gr.Blocks(theme=gr.themes.Soft(), title="Student Learning Assistant") as app:
        # Session state
        session_token = gr.State(value=generate_session_token())
        profile_manager.set_session(session_token.value)
        
        # Track completion status for each tab
        tab_completed = gr.State({
            0: False,  # Transcript Upload
            1: False,  # Learning Style Quiz
            2: False,  # Personal Questions
            3: False,  # Save & Review
            4: False   # AI Assistant
        })
        
        # Custom CSS with dark mode support
        app.css = """
        .gradio-container { max-width: 1200px !important; margin: 0 auto !important; }
        .tab-content { padding: 20px !important; border: 1px solid #e0e0e0 !important; border-radius: 8px !important; margin-top: 10px !important; }
        .completed-tab { background: #4CAF50 !important; color: white !important; }
        .incomplete-tab { background: #E0E0E0 !important; }
        .nav-message { padding: 10px; margin: 10px 0; border-radius: 4px; background-color: #ffebee; color: #c62828; }
        .file-upload { border: 2px dashed #4CAF50 !important; padding: 20px !important; border-radius: 8px !important; }
        .progress-bar { height: 5px; background: linear-gradient(to right, #4CAF50, #8BC34A); margin-bottom: 15px; border-radius: 3px; }
        .quiz-question { margin-bottom: 15px; padding: 15px; background: #f5f5f5; border-radius: 5px; }
        .quiz-results { margin-top: 20px; padding: 20px; background: #e8f5e9; border-radius: 8px; }
        .error-message { color: #d32f2f; background-color: #ffebee; padding: 10px; border-radius: 4px; margin: 10px 0; }
        
        /* Dark mode support */
        .dark .tab-content { background-color: #2d2d2d !important; border-color: #444 !important; }
        .dark .quiz-question { background-color: #3d3d3d !important; }
        .dark .quiz-results { background-color: #2e3d2e !important; }
        .dark textarea, .dark input { background-color: #333 !important; color: #eee !important; }
        .dark .output-markdown { color: #eee !important; }
        .dark .chatbot { background-color: #333 !important; }
        .dark .chatbot .user, .dark .chatbot .assistant { color: #eee !important; }
        """
        
        # Header with dark mode toggle
        with gr.Row():
            with gr.Column(scale=4):
                gr.Markdown("""
                # Student Learning Assistant
                **Your personalized education companion**  
                Complete each step to get customized learning recommendations.
                """)
            with gr.Column(scale=1):
                dark_mode = gr.Checkbox(label="Dark Mode", value=False)
        
        # Navigation buttons
        with gr.Row():
            with gr.Column(scale=1, min_width=100):
                step1 = gr.Button("1. Transcript", elem_classes="incomplete-tab")
            with gr.Column(scale=1, min_width=100):
                step2 = gr.Button("2. Quiz", elem_classes="incomplete-tab", interactive=False)
            with gr.Column(scale=1, min_width=100):
                step3 = gr.Button("3. Profile", elem_classes="incomplete-tab", interactive=False)
            with gr.Column(scale=1, min_width=100):
                step4 = gr.Button("4. Review", elem_classes="incomplete-tab", interactive=False)
            with gr.Column(scale=1, min_width=100):
                step5 = gr.Button("5. Assistant", elem_classes="incomplete-tab", interactive=False)

        nav_message = gr.HTML(visible=False)

        # Main tabs container - Now VISIBLE
        with gr.Tabs(visible=True) as tabs:
            # ===== TAB 1: TRANSCRIPT UPLOAD =====
            with gr.Tab("Transcript", id=0):
                with gr.Row():
                    with gr.Column(scale=1):
                        gr.Markdown("### Step 1: Upload Your Transcript")
                        with gr.Group(elem_classes="file-upload"):
                            file_input = gr.File(
                                label="Drag and drop your transcript here (PDF or Image)",
                                file_types=ALLOWED_FILE_TYPES,
                                type="filepath"
                            )
                            upload_btn = gr.Button("Analyze Transcript", variant="primary")
                            file_error = gr.HTML(visible=False)
                    
                    with gr.Column(scale=2):
                        transcript_output = gr.Textbox(
                            label="Analysis Results", 
                            lines=20,
                            interactive=False
                        )
                        transcript_data = gr.State()

                def process_transcript(file_obj, current_tab_status):
                    try:
                        if not file_obj:
                            raise ValueError("Please upload a transcript file first.")
                            
                        output_text, data = parse_transcript(file_obj)
                        
                        if "Error" in output_text:
                            return (
                                output_text,
                                None,
                                current_tab_status,
                                gr.update(),
                                gr.update(),
                                gr.update(visible=True, value=f"<div class='error-message'>{output_text}</div>"),
                                gr.update(visible=False)
                            )
                        
                        new_status = current_tab_status.copy()
                        new_status[0] = True
                        return (
                            output_text, 
                            data, 
                            new_status,
                            gr.update(elem_classes="completed-tab"),
                            gr.update(interactive=True),
                            gr.update(visible=False),
                            gr.update(visible=False)
                        )
                        
                    except Exception as e:
                        error_msg = f"Error processing transcript: {str(e)}"
                        if "PDF" in str(e):
                            error_msg += "\n\nTIPS:\n- Try converting to image (screenshot)\n- Ensure text is selectable in PDF\n- Try a different PDF reader"
                        return (
                            error_msg,
                            None,
                            current_tab_status,
                            gr.update(),
                            gr.update(),
                            gr.update(visible=True, value=f"<div class='error-message'>{error_msg}</div>"),
                            gr.update(visible=False)
                        )
                
                upload_btn.click(
                    process_transcript,
                    inputs=[file_input, tab_completed],
                    outputs=[transcript_output, transcript_data, tab_completed, step1, step2, file_error, nav_message]
                )

            # ===== TAB 2: LEARNING STYLE QUIZ =====
            with gr.Tab("Learning Style Quiz", id=1):
                with gr.Column():
                    gr.Markdown("### Step 2: Discover Your Learning Style")
                    progress = gr.HTML("<div class='progress-bar' style='width: 0%'></div>")
                    
                    quiz_components = []
                    with gr.Accordion("Quiz Questions", open=True):
                        for i, (question, options) in enumerate(zip(learning_style_quiz.questions, learning_style_quiz.options)):
                            with gr.Group(elem_classes="quiz-question"):
                                q = gr.Radio(
                                    options,
                                    label=f"{i+1}. {question}",
                                    show_label=True
                                )
                                quiz_components.append(q)
                    
                    with gr.Row():
                        quiz_submit = gr.Button("Submit Quiz", variant="primary")
                        quiz_clear = gr.Button("Clear Answers")
                    
                    quiz_alert = gr.HTML(visible=False)
                    learning_output = gr.Markdown(
                        label="Your Learning Style Results",
                        visible=False,
                        elem_classes="quiz-results"
                    )

                # Update progress bar as questions are answered
                for component in quiz_components:
                    component.change(
                        fn=lambda *answers: {
                            progress: gr.HTML(
                                f"<div class='progress-bar' style='width: {sum(1 for a in answers if a)/len(answers)*100}%'></div>"
                            )
                        },
                        inputs=quiz_components,
                        outputs=progress
                    )
                
                def submit_quiz_and_update(*args):
                    current_tab_status = args[0]
                    answers = args[1:]
                    
                    try:
                        result = learning_style_quiz.evaluate_quiz(*answers)
                        new_status = current_tab_status.copy()
                        new_status[1] = True
                        return (
                            result,
                            gr.update(visible=True),
                            new_status,
                            gr.update(elem_classes="completed-tab"),
                            gr.update(interactive=True),
                            gr.update(value="<div class='alert-box'>Quiz submitted successfully!</div>", visible=True),
                            gr.update(visible=False)
                        )
                    except Exception as e:
                        return (
                            f"Error evaluating quiz: {str(e)}",
                            gr.update(visible=True),
                            current_tab_status,
                            gr.update(),
                            gr.update(),
                            gr.update(value=f"<div class='error-message'>Error: {str(e)}</div>", visible=True),
                            gr.update(visible=False)
                        )
                
                quiz_submit.click(
                    fn=submit_quiz_and_update,
                    inputs=[tab_completed] + quiz_components,
                    outputs=[learning_output, learning_output, tab_completed, step2, step3, quiz_alert, nav_message]
                )
                
                quiz_clear.click(
                    fn=lambda: [None] * len(quiz_components),
                    outputs=quiz_components
                ).then(
                    fn=lambda: gr.HTML("<div class='progress-bar' style='width: 0%'></div>"),
                    outputs=progress
                )

            # ===== TAB 3: PERSONAL QUESTIONS =====
            with gr.Tab("Personal Profile", id=2):
                with gr.Row():
                    with gr.Column(scale=1):
                        gr.Markdown("### Step 3: Tell Us About Yourself")
                        with gr.Group():
                            name = gr.Textbox(label="Full Name", placeholder="Your name")
                            age = gr.Number(label="Age", minimum=MIN_AGE, maximum=MAX_AGE, precision=0)
                            interests = gr.Textbox(
                                label="Your Interests/Hobbies",
                                placeholder="e.g., Science, Music, Sports, Art..."
                            )
                        
                        save_personal_btn = gr.Button("Save Information", variant="primary")
                        save_confirmation = gr.HTML(visible=False)
                    
                    with gr.Column(scale=1):
                        gr.Markdown("### Favorites")
                        with gr.Group():
                            movie = gr.Textbox(label="Favorite Movie")
                            movie_reason = gr.Textbox(label="Why do you like it?", lines=2)
                            show = gr.Textbox(label="Favorite TV Show")
                            show_reason = gr.Textbox(label="Why do you like it?", lines=2)
                            book = gr.Textbox(label="Favorite Book")
                            book_reason = gr.Textbox(label="Why do you like it?", lines=2)
                            character = gr.Textbox(label="Favorite Character (from any story)")
                            character_reason = gr.Textbox(label="Why do you like them?", lines=2)
                        
                        # Added blog section
                        with gr.Accordion("Personal Blog (Optional)", open=False):
                            blog = gr.Textbox(
                                label="Share your thoughts",
                                placeholder="Write something about yourself, your goals, or anything you'd like to share...",
                                lines=5
                            )
                
                def save_personal_info(name, age, interests, current_tab_status):
                    try:
                        name = validate_name(name)
                        age = validate_age(age)
                        interests = sanitize_input(interests)
                        
                        new_status = current_tab_status.copy()
                        new_status[2] = True
                        return (
                            new_status,
                            gr.update(elem_classes="completed-tab"),
                            gr.update(interactive=True),
                            gr.update(value="<div class='alert-box'>Information saved!</div>", visible=True),
                            gr.update(visible=False)
                        )
                    except Exception as e:
                        return (
                            current_tab_status,
                            gr.update(),
                            gr.update(),
                            gr.update(visible=False),
                            gr.update(visible=True, value=f"<div class='error-message'>Error: {str(e)}</div>")
                        )
                
                save_personal_btn.click(
                    fn=save_personal_info,
                    inputs=[name, age, interests, tab_completed],
                    outputs=[tab_completed, step3, step4, save_confirmation, nav_message]
                )

            # ===== TAB 4: SAVE & REVIEW =====
            with gr.Tab("Save Profile", id=3):
                with gr.Row():
                    with gr.Column(scale=1):
                        gr.Markdown("### Step 4: Review & Save Your Profile")
                        with gr.Group():
                            load_profile_dropdown = gr.Dropdown(
                                label="Load Existing Profile",
                                choices=profile_manager.list_profiles(session_token.value),
                                visible=False
                            )
                            with gr.Row():
                                load_btn = gr.Button("Load", visible=False)
                                delete_btn = gr.Button("Delete", variant="stop", visible=False)
                        
                        save_btn = gr.Button("Save Profile", variant="primary")
                        clear_btn = gr.Button("Clear Form")
                    
                    with gr.Column(scale=2):
                        output_summary = gr.Markdown(
                            "Your profile summary will appear here after saving.",
                            label="Profile Summary"
                        )

                def save_profile_and_update(name, age, interests, transcript_data, learning_style, 
                                         movie, movie_reason, show, show_reason,
                                         book, book_reason, character, character_reason, blog,
                                         current_tab_status):
                    try:
                        summary = profile_manager.save_profile(
                            name, age, interests, transcript_data, learning_style,
                            movie, movie_reason, show, show_reason,
                            book, book_reason, character, character_reason, blog
                        )
                        new_status = current_tab_status.copy()
                        new_status[3] = True
                        return (
                            summary,
                            new_status,
                            gr.update(elem_classes="completed-tab"),
                            gr.update(interactive=True),
                            gr.update(visible=False)
                        )
                    except Exception as e:
                        return (
                            f"Error saving profile: {str(e)}",
                            current_tab_status,
                            gr.update(),
                            gr.update(),
                            gr.update(visible=True, value=f"<div class='error-message'>Error: {str(e)}</div>")
                        )
                
                save_btn.click(
                    fn=save_profile_and_update,
                    inputs=[
                        name, age, interests, transcript_data, learning_output,
                        movie, movie_reason, show, show_reason,
                        book, book_reason, character, character_reason, blog,
                        tab_completed
                    ],
                    outputs=[output_summary, tab_completed, step4, step5, nav_message]
                ).then(
                    fn=lambda: profile_manager.list_profiles(session_token.value),
                    outputs=load_profile_dropdown
                ).then(
                    fn=lambda: gr.update(visible=bool(profile_manager.list_profiles(session_token.value))),
                    outputs=load_btn
                ).then(
                    fn=lambda: gr.update(visible=bool(profile_manager.list_profiles(session_token.value))),
                    outputs=delete_btn
                )
                
                def delete_profile(name, session_token):
                    if not name:
                        raise gr.Error("Please select a profile to delete")
                    try:
                        profile_path = profile_manager.get_profile_path(name)
                        if profile_path.exists():
                            profile_path.unlink()
                        return "Profile deleted successfully", ""
                    except Exception as e:
                        raise gr.Error(f"Error deleting profile: {str(e)}")
                
                delete_btn.click(
                    fn=delete_profile,
                    inputs=[load_profile_dropdown, session_token],
                    outputs=[output_summary, load_profile_dropdown]
                ).then(
                    fn=lambda: profile_manager.list_profiles(session_token.value),
                    outputs=load_profile_dropdown
                ).then(
                    fn=lambda: gr.update(visible=bool(profile_manager.list_profiles(session_token.value))),
                    outputs=load_btn
                ).then(
                    fn=lambda: gr.update(visible=bool(profile_manager.list_profiles(session_token.value))),
                    outputs=delete_btn
                )
                
                clear_btn.click(
                    fn=lambda: [gr.update(value="") for _ in range(12)],
                    outputs=[
                        name, age, interests, 
                        movie, movie_reason, show, show_reason,
                        book, book_reason, character, character_reason,
                        output_summary
                    ]
                )

            # ===== TAB 5: AI ASSISTANT =====
            with gr.Tab("AI Assistant", id=4):
                gr.Markdown("## Your Personalized Learning Assistant")
                gr.Markdown("Ask me anything about studying, your courses, grades, or learning strategies.")
                
                # Create a wrapper function that properly awaits the async function
                async def chat_wrapper(message: str, history: List[List[str]]):
                    response = await teaching_assistant.generate_response(
                        message, 
                        history, 
                        session_token.value
                    )
                    return response
                
                chatbot = gr.ChatInterface(
                    fn=chat_wrapper,
                    examples=[
                        "How should I study for my next math test?",
                        "What's my current GPA?",
                        "Show me my course history",
                        "How can I improve my grades in science?",
                        "What study methods match my learning style?"
                    ],
                    title=""
                )

        # Navigation logic
        def navigate_to_tab(tab_index: int, tab_completed_status):
            current_tab = tabs.selected
            
            # Allow backward navigation
            if tab_index <= current_tab:
                return gr.Tabs(selected=tab_index), gr.update(visible=False)
            
            # Check if current tab is completed
            if not tab_completed_status.get(current_tab, False):
                messages = {
                    0: "Please complete the transcript analysis first.",
                    1: "Please complete the learning style quiz first.",
                    2: "Please fill out your personal information first.",
                    3: "Please save your profile first."
                }
                return (
                    gr.Tabs(selected=current_tab),
                    gr.update(
                        value=f"<div class='error-message'>⚠️ {messages.get(current_tab, 'Please complete this step first')}</div>", 
                        visible=True
                    )
                )
            
            return gr.Tabs(selected=tab_index), gr.update(visible=False)
        
        # Connect navigation buttons
        step1.click(
            lambda idx, status: navigate_to_tab(idx, status),
            inputs=[gr.State(0), tab_completed],
            outputs=[tabs, nav_message]
        )
        step2.click(
            lambda idx, status: navigate_to_tab(idx, status),
            inputs=[gr.State(1), tab_completed],
            outputs=[tabs, nav_message]
        )
        step3.click(
            lambda idx, status: navigate_to_tab(idx, status),
            inputs=[gr.State(2), tab_completed],
            outputs=[tabs, nav_message]
        )
        step4.click(
            lambda idx, status: navigate_to_tab(idx, status),
            inputs=[gr.State(3), tab_completed],
            outputs=[tabs, nav_message]
        )
        step5.click(
            lambda idx, status: navigate_to_tab(idx, status),
            inputs=[gr.State(4), tab_completed],
            outputs=[tabs, nav_message]
        )
        
        # Dark mode toggle
        def toggle_dark_mode(dark):
            return gr.themes.Soft(primary_hue="blue", secondary_hue="gray") if not dark else gr.themes.Soft(primary_hue="blue", secondary_hue="gray", neutral_hue="slate")
        
        dark_mode.change(
            fn=toggle_dark_mode,
            inputs=dark_mode,
            outputs=None
        )
        
        # Load model on startup
        app.load(fn=lambda: model_loader.load_model(), outputs=[])
    
    return app

# Create and launch the interface
app = create_interface()

if __name__ == "__main__":
    app.launch()