breakpilot-lehrer/backend-lehrer/correction_helpers.py

"""
Correction API - Helper functions for grading, feedback, and OCR processing.
"""

import logging
from typing import List, Dict

from correction_models import AnswerEvaluation, CorrectionStatus, Correction

logger = logging.getLogger(__name__)

# FileProcessor requires OpenCV with libGL - make optional for CI
try:
    from services.file_processor import FileProcessor, ProcessingResult
    _ocr_available = True
except (ImportError, OSError):
    FileProcessor = None  # type: ignore
    ProcessingResult = None  # type: ignore
    _ocr_available = False

# PDF service requires WeasyPrint with system libraries - make optional for CI
try:
    from services.pdf_service import PDFService, CorrectionData, StudentInfo
    _pdf_available = True
except (ImportError, OSError):
    PDFService = None  # type: ignore
    CorrectionData = None  # type: ignore
    StudentInfo = None  # type: ignore
    _pdf_available = False


# ============================================================================
# In-Memory Storage (spaeter durch DB ersetzen)
# ============================================================================

corrections_store: Dict[str, Correction] = {}


# ============================================================================
# Helper Functions
# ============================================================================

def calculate_grade(percentage: float) -> str:
    """Berechnet Note aus Prozent (deutsches System)."""
    if percentage >= 92:
        return "1"
    elif percentage >= 81:
        return "2"
    elif percentage >= 67:
        return "3"
    elif percentage >= 50:
        return "4"
    elif percentage >= 30:
        return "5"
    else:
        return "6"


def generate_ai_feedback(
    evaluations: List[AnswerEvaluation],
    total_points: float,
    max_points: float,
    subject: str
) -> str:
    """Generiert KI-Feedback basierend auf Bewertung."""
    # Ohne LLM: Einfaches Template-basiertes Feedback
    percentage = (total_points / max_points * 100) if max_points > 0 else 0
    correct_count = sum(1 for e in evaluations if e.is_correct)
    total_count = len(evaluations)

    if percentage >= 90:
        intro = "Hervorragende Leistung!"
    elif percentage >= 75:
        intro = "Gute Arbeit!"
    elif percentage >= 60:
        intro = "Insgesamt eine solide Leistung."
    elif percentage >= 50:
        intro = "Die Arbeit zeigt Grundkenntnisse, aber es gibt Verbesserungsbedarf."
    else:
        intro = "Es sind deutliche Wissensluecken erkennbar."

    # Finde Verbesserungsbereiche
    weak_areas = [e for e in evaluations if not e.is_correct]
    strengths = [e for e in evaluations if e.is_correct and e.confidence > 0.8]

    feedback_parts = [intro]

    if strengths:
        feedback_parts.append(
            f"Besonders gut geloest: Aufgabe(n) {', '.join(str(s.question_number) for s in strengths[:3])}."
        )

    if weak_areas:
        feedback_parts.append(
            f"Uebungsbedarf bei: Aufgabe(n) {', '.join(str(w.question_number) for w in weak_areas[:3])}."
        )

    feedback_parts.append(
        f"Ergebnis: {correct_count} von {total_count} Aufgaben korrekt ({percentage:.1f}%)."
    )

    return " ".join(feedback_parts)


async def process_ocr(correction_id: str, file_path: str):
    """Background Task fuer OCR-Verarbeitung."""
    from datetime import datetime

    correction = corrections_store.get(correction_id)
    if not correction:
        return

    try:
        correction.status = CorrectionStatus.PROCESSING
        corrections_store[correction_id] = correction

        # OCR durchfuehren
        processor = FileProcessor()
        result = processor.process_file(file_path)

        if result.success and result.text:
            correction.extracted_text = result.text
            correction.status = CorrectionStatus.OCR_COMPLETE
        else:
            correction.status = CorrectionStatus.ERROR

        correction.updated_at = datetime.utcnow()
        corrections_store[correction_id] = correction

    except Exception as e:
        logger.error(f"OCR error for {correction_id}: {e}")
        correction.status = CorrectionStatus.ERROR
        correction.updated_at = datetime.utcnow()
        corrections_store[correction_id] = correction