kcc_core/topic_guard.py

"""Cheap regex/keyword guard. Blocks non-agriculture before any FAISS/LLM cost."""
from __future__ import annotations
import re

_AGRI_SIGNALS = {
    # Crops
    "crop","crops","plant","plants","planting","seed","seeds","soil","fertilizer",
    "pesticide","pest","disease","farm","farmer","farming","kisan","kheti","fasal",
    "gehu","gehun","dhan","kapas","tamatar","aloo","pyaz","mandi","bhav","rate",
    "harvest","harvesting","sow","sown","sowing","grow","grows","growing","grown",
    "variety","varieties","cultivar","cultivars","season","seasonal","kharif","rabi","zaid",
    "irrigation","spray","insecticide","fungicide","organic","yield","yields","blight",
    "wheat","rice","paddy","cotton","maize","sugarcane","soybean","mustard",
    "chilli","brinjal","onion","tomato","potato","groundnut","gram","pulses","pulse",
    "barley","jowar","bajra","ragi","arhar","tur","moong","urad","masoor","chana",
    # Pests / diseases
    "aphid","borer","mildew","rust","wilt","thrips","mite","whitefly",
    "caterpillar","jassid","leaf spot","mosaic","virus","rot","fungus",
    "blast","armyworm","bollworm","helicoverpa","spodoptera","girdle",
    # Hindi / regional
    "khaad","dawai","beej","sinchai","pattiya","pattiyan","keeda","bimari",
    "rog","upchar","khet","paidavar","safed makhi","tela","mahu","tikda",
    # Weather for crops
    "pala","thand","frost","baarish","drought","sukha","heat stress",
    "ola","flood","andhi",
    # Schemes / advisory
    "kvk","icar","pm kisan","fasal bima","kcc","kisan call","advisory",
    # Chemicals (banned-chemical questions ARE agri)
    "endosulfan","monocrotophos","chlorpyrifos","imidacloprid","emamectin",
    "mancozeb","propiconazole","thiamethoxam","rhizobium","urea","dap",
    # Devanagari
    "फसल","खेत","किसान","खाद","बीज","धान","गेहूं","कपास",
    "मक्का","सरसों","चना","अरहर","मूंग","प्याज","आलू",
    "कीड़ा","बीमारी","दवाइ","स्प्रे","उपचार","सिंचाई",
}

_NON_AGRI_RE = re.compile(
    r"\b(stock market|share market|bitcoin|crypto|politics|election|"
    r"movie|cricket|football|recipe|cooking|exam|bank account|insurance claim|"
    r"marriage|divorce|job|salary|relationship|celebrity|news|tv show|"
    r"web series|ipl|bollywood|actor|actress|pakistan|china|war|army|"
    r"love|dating|girlfriend|boyfriend|password|hack|code)\b",
    re.IGNORECASE,
)

OFF_TOPIC_MSG = (
    "I can only help with agriculture-related questions — crops, pests, diseases, "
    "fertilizers, mandi prices, irrigation, seeds, and farming advice. "
    "Please ask a farming question and I'll be happy to help! 🌾\n\n"
    "Main sirf kheti-baadi, fasal, keede-makode, bimari, khaad, mandi bhav, "
    "aur kisan-sambandhit sawalon ka jawab de sakta hoon."
)


def is_agriculture_query(query: str) -> bool:
    """Return True if the query looks like an agriculture question.

    Short follow-ups (≤3 words) always pass — they're contextual replies.
    Devanagari script defaults to allow unless a hard non-agri regex hits.
    """
    stripped = query.strip()
    q_lower_short = stripped.lower()
    # ≤3 words is usually a follow-up like "ok", "haan", "thanks" — but only
    # bypass if no obvious non-agri keyword is present.
    if len(stripped.split()) <= 3 and not _NON_AGRI_RE.search(q_lower_short):
        return True

    if any('ऀ' <= c <= 'ॿ' for c in stripped):
        return not _NON_AGRI_RE.search(stripped.lower())

    q_lower = stripped.lower()
    if _NON_AGRI_RE.search(q_lower):
        words = set(re.findall(r"\b\w+\b", q_lower))
        return bool(words.intersection(_AGRI_SIGNALS))

    words = set(re.findall(r"\b\w+\b", q_lower))
    return bool(words.intersection(_AGRI_SIGNALS))