backend/services/report_generator.py

"""Utilities for assembling and exporting interview reports.

This module provides two primary helpers used by the recruiter dashboard:

``generate_llm_interview_report(application)``
    Given a candidate's ``Application`` record, assemble a plain‑text report
    summarising the interview.  Because the interview process currently
    executes entirely client‑side and does not persist questions or answers
    to the database, this report focuses on the information available on
    the server: the candidate's profile, the job requirements and a skills
    match score.  Should future iterations store richer interview data
    server‑side, this function can be extended to include question/answer
    transcripts, per‑question scores and LLM‑generated feedback.

``create_pdf_report(report_text)``
    Convert a multi‑line string into a simple PDF.  The implementation
    leverages Matplotlib's PDF backend (available by default) to avoid
    heavyweight dependencies such as ReportLab or WeasyPrint, which are
    absent from the runtime environment.  Text is wrapped and split
    across multiple pages as necessary.
"""

from __future__ import annotations
import json
from io import BytesIO
import textwrap
from typing import List, Dict, Any, Tuple
import matplotlib.pyplot as plt
from matplotlib.backends.backend_pdf import PdfPages
import matplotlib.patches as mpatches
from matplotlib.patches import Rectangle, FancyBboxPatch
from datetime import datetime


def generate_llm_interview_report(application) -> str:
    """Generate a human‑readable interview report for a candidate.
    
    The report includes the candidate's name and email, job details,
    application date, a computed skills match summary and placeholder
    sections for future enhancements.  If server‑side storage of
    question/answer pairs is added later, this function can be updated
    to incorporate those details.
    
    Parameters
    ----------
    application : backend.models.database.Application
        The SQLAlchemy Application instance representing the candidate's
        job application.  Assumed to have related ``job`` and
        ``date_applied`` attributes available.
    
    Returns
    -------
    str
        A multi‑line string containing the report contents.
    """
    # Defensive imports to avoid circular dependencies at import time
    try:
        from datetime import datetime  # noqa: F401
    except Exception:
        pass

    # Extract candidate skills and job skills
    try:
        candidate_features = json.loads(application.extracted_features) if application.extracted_features else {}
    except Exception:
        candidate_features = {}

    candidate_skills: List[str] = candidate_features.get('skills', []) or []
    job_skills: List[str] = []
    try:
        job_skills = json.loads(application.job.skills) if application.job and application.job.skills else []
    except Exception:
        job_skills = []

    # Compute skills match ratio and label.  Normalise to lower case for
    # comparison and avoid dividing by zero when ``job_skills`` is empty.
    candidate_set = {s.strip().lower() for s in candidate_skills}
    job_set = {s.strip().lower() for s in job_skills}
    common = candidate_set & job_set
    ratio = len(common) / len(job_set) if job_set else 0.0

    if ratio >= 0.75:
        score_label = 'Excellent'
    elif ratio >= 0.5:
        score_label = 'Good'
    elif ratio >= 0.25:
        score_label = 'Medium'
    else:
        score_label = 'Poor'

    # Assemble report lines
    lines: List[str] = []
    lines.append('Interview Report')
    lines.append('=================')
    lines.append('')
    lines.append(f'Candidate Name: {application.name}')
    lines.append(f'Candidate Email: {application.email}')
    if application.job:
        lines.append(f'Job Applied: {application.job.role}')
        lines.append(f'Company: {application.job.company}')
    else:
        lines.append('Job Applied: N/A')
        lines.append('Company: N/A')

    # Format date_applied if available
    try:
        date_str = application.date_applied.strftime('%Y-%m-%d') if application.date_applied else 'N/A'
    except Exception:
        date_str = 'N/A'
    lines.append(f'Date Applied: {date_str}')
    lines.append('')

    lines.append('Skills Match Summary:')
    # Represent required and candidate skills as comma‑separated lists.  Use
    # title‑case for presentation and handle empty lists gracefully.
    formatted_job_skills = ', '.join(job_skills) if job_skills else 'N/A'
    formatted_candidate_skills = ', '.join(candidate_skills) if candidate_skills else 'N/A'
    formatted_common = ', '.join(sorted(common)) if common else 'None'

    lines.append(f'  Required Skills: {formatted_job_skills}')
    lines.append(f'  Candidate Skills: {formatted_candidate_skills}')
    lines.append(f'  Skills in Common: {formatted_common}')
    lines.append(f'  Match Ratio: {ratio * 100:.0f}%')
    lines.append(f'  Score: {score_label}')
    lines.append('')

    lines.append('Interview Transcript & Evaluation:')
    try:
        if application.interview_log:
            try:
                qa_log = json.loads(application.interview_log)
            except Exception:
                qa_log = []
            
            if qa_log:
                for idx, entry in enumerate(qa_log, 1):
                    q = entry.get("question", "N/A")
                    a = entry.get("answer", "N/A")
                    
                    # Handle salary question specifically
                    if "salary" in q.lower() and (a == "0$" or a == "0" or a == "$0"):
                        a = "Prefer not to disclose"
                    
                    eval_score = entry.get("evaluation", {}).get("score", "N/A")
                    eval_feedback = entry.get("evaluation", {}).get("feedback", "N/A")
                    
                    lines.append(f"\nQuestion {idx}: {q}")
                    lines.append(f"Answer: {a}")
                    lines.append(f"Score: {eval_score}")
                    lines.append(f"Feedback: {eval_feedback}")
            else:
                lines.append("No interview log data recorded.")
        else:
            lines.append("No interview log data recorded.")
    except Exception as e:
        lines.append(f"Error loading interview log: {e}")

    return '\n'.join(lines)


def create_pdf_report(report_text: str) -> BytesIO:
    """Convert a formatted report into a clean, professional A4 PDF."""
    buffer = BytesIO()
    
    # A4 dimensions in inches (210mm x 297mm)
    A4_WIDTH = 8.27
    A4_HEIGHT = 11.69
    
    # Margins in inches
    LEFT_MARGIN = 0.75
    RIGHT_MARGIN = 0.75
    TOP_MARGIN = 0.75
    BOTTOM_MARGIN = 0.75
    
    # Calculate content area
    CONTENT_WIDTH = A4_WIDTH - LEFT_MARGIN - RIGHT_MARGIN
    CONTENT_HEIGHT = A4_HEIGHT - TOP_MARGIN - BOTTOM_MARGIN
    
    # Professional color scheme - single accent color
    ACCENT_COLOR = '#1e40af'  # Dark blue
    TEXT_COLOR = '#111827'    # Dark gray/black
    LIGHT_GRAY = '#f8fafc'    # Light background
    BORDER_COLOR = '#e2e8f0'  # Light border
    
    # Parse report data
    report_data = _parse_report_text(report_text)
    
    # Create PDF
    with PdfPages(buffer) as pdf:
        # Page 1: Header, Candidate Info, and Skills Summary
        fig = plt.figure(figsize=(A4_WIDTH, A4_HEIGHT))
        fig.patch.set_facecolor('white')
        
        # Create main axis
        ax = fig.add_subplot(111)
        ax.set_xlim(0, A4_WIDTH)
        ax.set_ylim(0, A4_HEIGHT)
        ax.axis('off')
        
        # Current Y position (start from top)
        y_pos = A4_HEIGHT - TOP_MARGIN
        
        # === HEADER SECTION ===
        # Clean header with company info
        ax.text(LEFT_MARGIN, y_pos, 'INTERVIEW ASSESSMENT REPORT', 
                fontsize=20, fontweight='bold', color=ACCENT_COLOR, fontfamily='sans-serif')
        
        # Date
        current_date = datetime.now().strftime('%B %d, %Y')
        ax.text(A4_WIDTH - RIGHT_MARGIN, y_pos, current_date,
                fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif',
                horizontalalignment='right')
        
        y_pos -= 0.8
        
        # === CANDIDATE INFO AND OVERALL SCORE ===
        # Large overall score box (prominent)
        overall_score = _calculate_overall_score(report_data)
        score_color = _get_score_color(overall_score['label'])
        
        # Score box on the right
        score_box_width = 2.5
        score_box_height = 1.8
        score_x = A4_WIDTH - RIGHT_MARGIN - score_box_width
        
        # Score background
        score_rect = FancyBboxPatch(
            (score_x, y_pos - score_box_height), score_box_width, score_box_height,
            boxstyle="round,pad=0.1",
            facecolor=LIGHT_GRAY,
            edgecolor=ACCENT_COLOR,
            linewidth=2
        )
        ax.add_patch(score_rect)
        
        # Large score percentage
        ax.text(score_x + score_box_width/2, y_pos - 0.6, f"{overall_score['percentage']:.0f}%",
                fontsize=32, fontweight='bold', color=ACCENT_COLOR, fontfamily='sans-serif',
                horizontalalignment='center', verticalalignment='center')
        
        # Score label
        ax.text(score_x + score_box_width/2, y_pos - 1.2, 'OVERALL SCORE',
                fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif',
                horizontalalignment='center', fontweight='bold')
        
        ax.text(score_x + score_box_width/2, y_pos - 1.5, overall_score['label'].upper(),
                fontsize=14, fontweight='bold', color=score_color, fontfamily='sans-serif',
                horizontalalignment='center')
        
        # Candidate information on the left
        info_width = CONTENT_WIDTH - score_box_width - 0.5
        
        # Candidate name (large)
        ax.text(LEFT_MARGIN, y_pos - 0.3, report_data['candidate_name'],
                fontsize=18, fontweight='bold', color=TEXT_COLOR, fontfamily='sans-serif')
        
        # Position and company
        ax.text(LEFT_MARGIN, y_pos - 0.7, f"{report_data['job_role']} • {report_data['company']}",
                fontsize=12, color=TEXT_COLOR, fontfamily='sans-serif')
        
        # Email and date
        ax.text(LEFT_MARGIN, y_pos - 1.0, f"Email: {report_data['candidate_email']}",
                fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif')
        
        ax.text(LEFT_MARGIN, y_pos - 1.3, f"Application Date: {report_data['date_applied']}",
                fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif')
        
        y_pos -= 2.5
        
        # === SKILLS MATCH SUMMARY ===
        # Section header
        ax.text(LEFT_MARGIN, y_pos, 'SKILLS MATCH SUMMARY',
                fontsize=14, fontweight='bold', color=ACCENT_COLOR, fontfamily='sans-serif')
        
        # Underline
        ax.plot([LEFT_MARGIN, LEFT_MARGIN + 3], [y_pos - 0.1, y_pos - 0.1], 
                color=ACCENT_COLOR, linewidth=2)
        
        y_pos -= 0.5
        
        skills_data = report_data['skills_match']
        
        # Skills match percentage bar
        bar_width = CONTENT_WIDTH - 1
        bar_height = 0.3
        
        # Background bar
        bg_rect = Rectangle((LEFT_MARGIN + 0.5, y_pos - bar_height), bar_width, bar_height,
                           facecolor=LIGHT_GRAY, edgecolor=BORDER_COLOR)
        ax.add_patch(bg_rect)
        
        # Progress bar
        progress_width = bar_width * (skills_data['ratio'] / 100)
        progress_rect = Rectangle((LEFT_MARGIN + 0.5, y_pos - bar_height), progress_width, bar_height,
                                 facecolor=ACCENT_COLOR, edgecolor='none')
        ax.add_patch(progress_rect)
        
        # Percentage text
        ax.text(LEFT_MARGIN + 0.5 + bar_width/2, y_pos - bar_height/2, 
                f"{skills_data['ratio']:.0f}% Skills Match",
                fontsize=11, fontweight='bold', color='white', fontfamily='sans-serif',
                horizontalalignment='center', verticalalignment='center')
        
        y_pos -= 0.8
        
        # Skills details (simplified)
        required_text = f"Required Skills: {skills_data['required']}"
        for line in textwrap.wrap(required_text, width=85):
            ax.text(LEFT_MARGIN, y_pos, line,
                    fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif')
            y_pos -= 0.3
        
        y_pos -= 0.3
        
        candidate_text = f"Candidate Skills: {skills_data['candidate']}"
        for line in textwrap.wrap(candidate_text, width=85):
            ax.text(LEFT_MARGIN, y_pos, line,
                    fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif')
            y_pos -= 0.3

        
        y_pos -= 0.3
        
        matching_text = f"Matching Skills: {skills_data['common']}"
        for line in textwrap.wrap(matching_text, width=85):
            ax.text(LEFT_MARGIN, y_pos, line,
                    fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif')
            y_pos -= 0.3

        
        y_pos -= 0.8
        
        # === INTERVIEW TRANSCRIPT PREVIEW ===
        if report_data['qa_log']:
            ax.text(LEFT_MARGIN, y_pos, 'INTERVIEW TRANSCRIPT',
                    fontsize=14, fontweight='bold', color=ACCENT_COLOR, fontfamily='sans-serif')
            
            # Underline
            ax.plot([LEFT_MARGIN, LEFT_MARGIN + 3], [y_pos - 0.1, y_pos - 0.1], 
                    color=ACCENT_COLOR, linewidth=2)
            
            y_pos -= 0.5
            
            # Show first 2-3 Q&As on first page
            max_qa_on_page1 = min(3, len(report_data['qa_log']))
            
            for i in range(max_qa_on_page1):
                qa = report_data['qa_log'][i]
                
                # Check if we have space
                if y_pos < BOTTOM_MARGIN + 2.2:
                    break
                
                # Question
                question_text = f"Q{i+1}: {qa['question']}"
                for line in textwrap.wrap(question_text, width=85):
                    ax.text(LEFT_MARGIN, y_pos, line,
                            fontsize=11, fontweight='bold', color=ACCENT_COLOR, fontfamily='sans-serif')
                    y_pos -= 0.25
                y_pos -= 0.15  # extra spacing after question block

                
                # Answer
                answer_text = qa['answer']
                if "salary" in qa['question'].lower() and (answer_text == "0$" or answer_text == "0" or answer_text == "$0"):
                    answer_text = "Prefer not to disclose"
                
                wrapped_answer = textwrap.fill(answer_text, width=85)
                answer_lines = wrapped_answer.split('\n')[:2]  # Max 2 lines
                
                for line in answer_lines:
                    ax.text(LEFT_MARGIN + 0.3, y_pos, line,
                            fontsize=10, color=TEXT_COLOR, fontfamily='sans-serif')
                    y_pos -= 0.25
                
                # Evaluation
                eval_color = _get_score_color(qa['score'])
                ax.text(LEFT_MARGIN + 0.3, y_pos, f"Evaluation: {qa['score']}",
                        fontsize=10, fontweight='bold', color=eval_color, fontfamily='sans-serif')
                
                y_pos -= 0.6
        
        # Save first page
        pdf.savefig(fig, bbox_inches='tight', pad_inches=0)
        plt.close(fig)
        
        # === PAGE 2: REMAINING TRANSCRIPT ===
        if report_data['qa_log'] and len(report_data['qa_log']) > max_qa_on_page1:
            _create_transcript_page(
                pdf,
                report_data['qa_log'][max_qa_on_page1:],  # Continue from the next unanswered question
                A4_WIDTH, A4_HEIGHT,
                LEFT_MARGIN, RIGHT_MARGIN, TOP_MARGIN, BOTTOM_MARGIN,
                ACCENT_COLOR, TEXT_COLOR,
                start_index=max_qa_on_page1 + 1  # Correct numbering
            )

    
    buffer.seek(0)
    return buffer


def _parse_report_text(report_text: str) -> Dict[str, Any]:
    """Parse the text report into structured data."""
    lines = report_text.split('\n')
    data = {
        'candidate_name': 'N/A',
        'candidate_email': 'N/A',
        'job_role': 'N/A',
        'company': 'N/A',
        'date_applied': 'N/A',
        'skills_match': {
            'required': 'N/A',
            'candidate': 'N/A',
            'common': 'N/A',
            'ratio': 0,
            'score': 'N/A'
        },
        'qa_log': []
    }
    
    current_question = None
    
    for line in lines:
        line = line.strip()
        if line.startswith('Candidate Name:'):
            data['candidate_name'] = line.split(':', 1)[1].strip()
        elif line.startswith('Candidate Email:'):
            data['candidate_email'] = line.split(':', 1)[1].strip()
        elif line.startswith('Job Applied:'):
            data['job_role'] = line.split(':', 1)[1].strip()
        elif line.startswith('Company:'):
            data['company'] = line.split(':', 1)[1].strip()
        elif line.startswith('Date Applied:'):
            data['date_applied'] = line.split(':', 1)[1].strip()
        elif line.startswith('Required Skills:'):
            data['skills_match']['required'] = line.split(':', 1)[1].strip()
        elif line.startswith('Candidate Skills:'):
            data['skills_match']['candidate'] = line.split(':', 1)[1].strip()
        elif line.startswith('Skills in Common:'):
            data['skills_match']['common'] = line.split(':', 1)[1].strip()
        elif line.startswith('Match Ratio:'):
            try:
                data['skills_match']['ratio'] = float(line.split(':')[1].strip().rstrip('%'))
            except:
                data['skills_match']['ratio'] = 0
        elif line.startswith('Score:') and 'skills_match' in str(data):
            data['skills_match']['score'] = line.split(':', 1)[1].strip()
        elif line.startswith('Question'):
            if current_question:
                data['qa_log'].append(current_question)
            current_question = {
                'question': line.split(':', 1)[1].strip() if ':' in line else line,
                'answer': '',
                'score': '',
                'feedback': ''
            }
        elif line.startswith('Answer:') and current_question:
            current_question['answer'] = line.split(':', 1)[1].strip()
        elif line.startswith('Score:') and current_question:
            current_question['score'] = line.split(':', 1)[1].strip()
        elif line.startswith('Feedback:') and current_question:
            current_question['feedback'] = line.split(':', 1)[1].strip()
    
    if current_question:
        data['qa_log'].append(current_question)
    
    return data


def _calculate_overall_score(report_data: Dict[str, Any]) -> Dict[str, Any]:
    """Calculate overall score from skills match and QA scores."""
    # Skills match contributes 40%
    skills_ratio = report_data['skills_match']['ratio'] / 100
    
    # QA scores contribute 60%
    qa_scores = []
    for qa in report_data['qa_log']:
        score_text = qa['score'].lower()
        if 'excellent' in score_text or '5' in score_text or '10' in score_text:
            qa_scores.append(1.0)
        elif 'good' in score_text or '4' in score_text or '8' in score_text or '9' in score_text:
            qa_scores.append(0.8)
        elif 'satisfactory' in score_text or 'medium' in score_text or '3' in score_text or '6' in score_text or '7' in score_text:
            qa_scores.append(0.6)
        elif 'needs improvement' in score_text or 'poor' in score_text or '2' in score_text or '4' in score_text or '5' in score_text:
            qa_scores.append(0.4)
        else:
            qa_scores.append(0.2)
    
    qa_average = sum(qa_scores) / len(qa_scores) if qa_scores else 0.5
    
    # Calculate weighted average
    overall = (skills_ratio * 0.4) + (qa_average * 0.6)
    percentage = overall * 100
    
    if overall >= 0.8:
        label = 'Excellent'
    elif overall >= 0.65:
        label = 'Good'
    elif overall >= 0.45:
        label = 'Satisfactory'
    else:
        label = 'Needs Improvement'
    
    return {'percentage': percentage, 'label': label}


def _get_score_color(score_label: str) -> str:
    """Get color based on score label."""
    score_label = score_label.lower()
    if 'excellent' in score_label:
        return '#059669'  # Green
    elif 'good' in score_label:
        return '#2563eb'  # Blue
    elif 'medium' in score_label or 'satisfactory' in score_label:
        return '#d97706'  # Orange
    else:
        return '#dc2626'  # Red


def _create_transcript_page(pdf, qa_log: List[Dict], page_width: float, page_height: float,
                          left_margin: float, right_margin: float, top_margin: float, bottom_margin: float,
                          accent_color: str, text_color: str, start_index: int = 1):
    """Create a clean page for remaining interview transcript."""
    content_width = page_width - left_margin - right_margin
    
    fig = plt.figure(figsize=(page_width, page_height))
    fig.patch.set_facecolor('white')
    ax = fig.add_subplot(111)
    ax.set_xlim(0, page_width)
    ax.set_ylim(0, page_height)
    ax.axis('off')
    
    # Start from top
    y_pos = page_height - top_margin
    
    # Page header
    ax.text(left_margin, y_pos, 'INTERVIEW TRANSCRIPT (CONTINUED)',
            fontsize=14, fontweight='bold', color=accent_color, fontfamily='sans-serif')
    
    # Underline
    ax.plot([left_margin, left_margin + 4], [y_pos - 0.1, y_pos - 0.1], 
            color=accent_color, linewidth=2)
    
    y_pos -= 0.8
    
    # Process remaining Q&As
    for i, qa in enumerate(qa_log):
        # Check if we have space for this Q&A
        if y_pos < bottom_margin + 1.5:
            break
        
        # Question
        question_text = f"Q{start_index + i}: {qa['question']}"
        wrapped_question = textwrap.fill(question_text, width=85)
        question_lines = wrapped_question.split('\n')
        
        for line in question_lines:
            ax.text(left_margin, y_pos, line,
                    fontsize=11, fontweight='bold', color=accent_color, fontfamily='sans-serif')
            y_pos -= 0.3
        
        y_pos -= 0.1
        
        # Answer
        answer_text = qa['answer']
        if "salary" in qa['question'].lower() and (answer_text == "0$" or answer_text == "0" or answer_text == "$0"):
            answer_text = "Prefer not to disclose"
        
        wrapped_answer = textwrap.fill(answer_text, width=80)
        answer_lines = wrapped_answer.split('\n')
        
        for line in answer_lines[:3]:  # Max 3 lines per answer
            ax.text(left_margin + 0.3, y_pos, line,
                    fontsize=10, color=text_color, fontfamily='sans-serif')
            y_pos -= 0.25
        
        # Evaluation
        eval_color = _get_score_color(qa['score'])
        ax.text(left_margin + 0.3, y_pos, f"Evaluation: {qa['score']}",
                fontsize=10, fontweight='bold', color=eval_color, fontfamily='sans-serif')
        
        y_pos -= 0.2
        
        # Feedback (if available and space permits)
        if qa['feedback'] and qa['feedback'] != 'N/A' and y_pos > bottom_margin + 0.8:
            feedback_text = f"Feedback: {qa['feedback']}"
            wrapped_feedback = textwrap.fill(feedback_text, width=75)
            feedback_lines = wrapped_feedback.split('\n')[:2]  # Max 2 lines
            
            for line in feedback_lines:
                ax.text(left_margin + 0.3, y_pos, line,
                        fontsize=9, color='#6b7280', fontfamily='sans-serif', style='italic')
                y_pos -= 0.2
        
        y_pos -= 0.4
        
        # Add separator line between questions
        if i < len(qa_log) - 1 and y_pos > bottom_margin + 1:
            ax.plot([left_margin + 0.5, left_margin + content_width - 0.5], 
                   [y_pos + 0.1, y_pos + 0.1], 
                   color='#e5e7eb', linewidth=0.5, linestyle='--')
            y_pos -= 0.3
    
    # Save page
    pdf.savefig(fig, bbox_inches='tight', pad_inches=0)
    plt.close(fig)


__all__ = ['generate_llm_interview_report', 'create_pdf_report']