data/qa_index.py

# data/qa_index.py – QA index from dataset + manual answers

from typing import Dict, Optional

from .loader import ENTRIES
from .text_utils import normalize_question
from .manual_answers import MANUAL_ANSWERS

QA_INDEX: Dict[str, str] = {}


def _build_qa_index():
    for obj in ENTRIES:
        for pair in obj.get("qa", []):
            q = pair.get("q", "")
            a = pair.get("a", "")
            if q and a:
                norm_q = normalize_question(q)
                QA_INDEX[norm_q] = a.strip()


_build_qa_index()


def answer_from_qa(question: str) -> Optional[str]:
    """
    1) Manual teacher answers (MANUAL_ANSWERS) – highest priority.
    2) Exact match from dataset QA_INDEX.
    3) Fuzzy match over dataset QA_INDEX.
    """
    norm_q = normalize_question(question)

    # 1) manual perfect answers
    if norm_q in MANUAL_ANSWERS:
        return MANUAL_ANSWERS[norm_q]

    # 2) exact match from dataset
    if norm_q in QA_INDEX:
        return QA_INDEX[norm_q]

    # 3) fuzzy match over dataset
    if not QA_INDEX:
        return None

    best_score = 0
    best_answer: Optional[str] = None

    for stored_q, a in QA_INDEX.items():
        score = 0
        for ch in norm_q:
            if ch and ch in stored_q:
                score += 1

        if score > best_score:
            best_score = score
            best_answer = a

    if best_score > 0:
        return best_answer

    return None