"""Detect the language of a customer message, so the model is told explicitly how to reply.

A general rule in the system prompt ("answer in the customer's language") is not enough on its own:
models drift, for example by ending an English reply with a Darija sentence. A short per-message
instruction based on this detection keeps every reply in the right language.
"""

from __future__ import annotations

import re

_ARABIC_LETTER = re.compile(r"[ء-ي]")
_LATIN_LETTER = re.compile(r"[A-Za-zÀ-ÿ]")

# Words that only appear in Algerian Darija, written in Arabic letters.
_DARIJA_ARABIC = re.compile(r"(قداش|قداه|واش|كاين|ماكانش|راني|راهو|راهي|نحب|تحب|بزاف|دروك|درك|تاع|باش|شحال|وين|علاش|كيفاش|بصح)")
# Darija written in Latin letters: common words, or digits used as letters (3, 7, 9) inside a word.
_DARIJA_LATIN_WORDS = re.compile(
    r"\b(ta3|wach|wesh|kayen|kayn|makach|rani|rak|nheb|t7eb|ch7al|chhal|bezzaf|chwiya|dork|drk|win|wech|saha|sahit|"
    r"salam|khoya|kifach|3lach|bach|mlih|ghir|rah|nbadel|nbeddel|3andkom|3andek|mta3)\b",
    re.IGNORECASE,
)
_DIGIT_AS_LETTER = re.compile(r"[a-z][379][a-z]|\b[379][a-z]{2,}|[a-z]{2,}[379]\b", re.IGNORECASE)
_FRENCH = re.compile(
    r"\b(je|vous|votre|vos|nous|est|les|des|une|pour|avec|merci|bonjour|bonsoir|commande|où|combien|livraison|"
    r"chiffres|derniers|ma|mon|mes|le|la|du|se|par|pas|quand|comment)\b",
    re.IGNORECASE,
)
_ENGLISH = re.compile(
    r"\b(the|you|your|is|are|and|can|could|order|please|with|what|where|when|how|my|do|does|i|me|to|of|all|list|who)\b",
    re.IGNORECASE,
)

DARIJA_LATIN = "Algerian Darija in Latin letters (Algerian words only, never Moroccan forms)"
DARIJA_ARABIC = "Algerian Darija in Arabic letters"
MSA = "Modern Standard Arabic (الفصحى), with no Darija words, including in the closing sentence"
FRENCH = "French only, with no Darija words, including in the closing sentence"
ENGLISH = "English only, with no Darija words, including in the closing sentence"


def detect_reply_language(text: str) -> str | None:
    """Return a short description of the language to reply in, or None when it is unclear."""
    arabic = len(_ARABIC_LETTER.findall(text))
    latin = len(_LATIN_LETTER.findall(text))
    if arabic == 0 and latin == 0:
        return None
    if arabic > latin:
        return DARIJA_ARABIC if _DARIJA_ARABIC.search(text) else MSA

    if _DARIJA_LATIN_WORDS.search(text) or _DIGIT_AS_LETTER.search(text):
        return DARIJA_LATIN
    french = len(_FRENCH.findall(text))
    english = len(_ENGLISH.findall(text))
    if french > english:
        return FRENCH
    if english > french:
        return ENGLISH
    return None


def language_instruction(text: str) -> str | None:
    language = detect_reply_language(text)
    if language is None:
        return None
    return f"Reply to the next customer message in {language}."