src/utils/metrics.py

import re
from typing import List, Dict, Tuple


def tokenize(text: str) -> List[str]:
    """Tokenize a string into lowercase words >3 chars."""
    return [w.lower() for w in re.findall(r"\w+", text) if len(w) > 3]


def compute_quality_scores(
        question: str,
        answer: str,
        sources: List[Dict],
) -> Tuple[float, float]:
    """Compute retrieval quality metrics (coverage & grounding).

    Args:
        question: User's question text.
        answer: Model-generated answer text.
        sources: Retrieved documents/chunks, each with a 'content' field.

    Returns:
        (coverage, grounding) as floats in [0.0, 1.0].
    """
    all_chunk_text = " ".join(s.get("content", "") for s in sources)
    q_tokens = tokenize(question)
    a_tokens = tokenize(answer)
    c_tokens = set(tokenize(all_chunk_text))

    if not c_tokens:
        return 0.0, 0.0

    coverage = (
        sum(1 for t in q_tokens if t in c_tokens) / len(q_tokens)
        if q_tokens
        else 0.0
    )
    grounding = (
        sum(1 for t in a_tokens if t in c_tokens) / len(a_tokens)
        if a_tokens
        else 0.0
    )

    return coverage, grounding