"""Lightweight, best-effort regex extractors for specific signal types the
severity model treats specially (pricing, leadership). These are heuristics,
not NLP - they exist to catch the common "$X/month" and "named a new CEO"
phrasings, not to parse arbitrary text reliably. Phase 7's LLM extraction
task is the higher-fidelity version of this; these run cheaply and
deterministically as part of scoring, without a model call.
"""

from __future__ import annotations

import re

_PRICE_RE = re.compile(r"\$\s?\d[\d,]*(?:\.\d{2})?\s*(?:/\s*(?:month|mo|year|yr))?")
_LEADERSHIP_TITLE_RE = re.compile(
    r"(?i)\b(Chief Executive Officer|CEO|Chief Financial Officer|CFO|Chief Technology Officer|"
    r"CTO|President|Chairman|Chairwoman|Chairperson)\b"
)


def extract_prices(text: str) -> set[str]:
    return set(_PRICE_RE.findall(text or ""))


def mentions_leadership_title(text: str) -> bool:
    return bool(_LEADERSHIP_TITLE_RE.search(text or ""))