scripts/convert_wisemoor_to_training_candidates.py

#!/usr/bin/env python3
"""Convert sanitized WiseMoor formulas into PINO training candidates.

This is intentionally gated: sanitized WiseMoor rows are name/amount formulas,
while the core PINO training set needs CAS-resolved, normalized formulas. The
script writes only formulas that meet configurable resolution thresholds and
also writes an audit file for the unresolved materials.
"""

from __future__ import annotations

import argparse
import json
import re
import sys
import unicodedata
from collections import Counter, defaultdict
from pathlib import Path
from typing import Any


ROOT = Path(__file__).resolve().parents[1]
sys.path.insert(0, str(ROOT / "scripts"))

from ingest_tgsc import build_tgsc_mapping, map_ingredient  # noqa: E402
from pino.registry import AromaRegistry  # noqa: E402


WISEMOOR_NAME_OVERRIDES: dict[str, str] = {
    # Solvents / carriers.
    "dep": "84-66-2",  # diethyl phthalate
    "d.p.g": "25265-71-8",
    "dpg": "25265-71-8",
    "dipropylene glycol": "25265-71-8",
    "t.e.c": "77-93-0",  # triethyl citrate
    "i.p.m": "110-27-0",  # isopropyl myristate
    # WiseMoor spelling/order variants for common aroma chemicals.
    "acetate benzyl": "140-11-4",
    "acetate bornyl": "76-49-3",
    "acetate bornyl l": "76-49-3",
    "acetate butyl": "123-86-4",
    "acetate citronellyl": "150-84-5",
    "acetate d.m.b.c": "151-05-3",
    "dmbc acetate": "151-05-3",
    "acetate ethyl": "141-78-6",
    "acetate geranyl": "105-87-3",
    "acetate hexenyl cis-3": "3681-71-8",
    "acetate hexyl": "142-92-7",
    "acetate isoamyl": "123-92-2",
    "acetate amyl": "628-63-7",
    "amyl acetate": "628-63-7",
    "acetate cinnamic": "103-54-8",
    "acetate linalyl": "115-95-7",
    "acetate linallyl": "115-95-7",
    "acetate neryl": "141-12-8",
    "acetate octyl": "112-14-1",
    "acetate prenyl": "1191-16-8",
    "acetate styrallyl": "93-92-5",
    "acetate styralyl": "93-92-5",
    "acetate terpenyl": "80-26-2",
    "acetate terpene": "80-26-2",
    "acetophenone": "98-86-2",
    "aldehyde alpha hexyl cinnamic": "101-86-0",
    "aldehyde alpha amyl cinnamic": "122-40-7",
    "aldehyde anisic": "123-11-5",
    "aldehyde c 08": "124-13-0",
    "aldehyde c 09": "112-31-2",
    "aldehyde c 11 und": "112-45-8",
    "aldehyde c 18": "104-61-0",
    "aldehydec 14": "104-67-6",
    "aldehyde cinammic": "104-55-2",
    "aldehyde cinnamic": "104-55-2",
    "aldehyde cyclamen": "103-95-7",
    "alcohol c-06": "111-27-3",
    "alcohol cinnamic": "104-54-1",
    "alcool cinnamique": "104-54-1",
    "alcohol isoamyl": "123-51-3",
    "alcohol phenylethyl": "60-12-8",
    "alcool phenylpropylique": "122-97-4",
    "alyl amyl glycolate": "67634-20-2",
    "allyl amyl glycolate": "67634-20-2",
    "anethol": "104-46-1",
    "anisic aldehyde": "123-11-5",
    "anthranilate dimethyl": "85-74-5",
    "anthranilate methyl": "134-20-3",
    "benzyl propionate": "122-63-4",
    "benzyl cinnamate": "103-41-3",
    "benzylic alcohol": "100-51-6",
    "benzophenone": "119-61-9",
    "benzoat benzyl": "120-51-4",
    "benzoate benzyl": "120-51-4",
    "benzoate methyle": "93-58-3",
    "bht": "128-37-0",
    "butyl quinoleine sec": "65442-31-1",
    "c 7 aldehyde": "111-71-7",
    "c 8 aldehyde": "124-13-0",
    "c 10 aldehyde": "112-31-2",
    "c 11 aldehyde": "112-44-7",
    "c 12 l aldehyde": "112-54-9",
    "c 12 mna": "110-41-8",
    "c 12 mna aldehyde": "110-41-8",
    "c 14 aldehyde": "104-67-6",
    "c 16 aldehyde": "77-83-8",
    "c 18 aldehyde": "104-61-0",
    "calone": "28940-11-6",
    "calone 1951": "28940-11-6",
    "cedramber": "67874-81-1",
    "cedrenol": "28231-03-0",
    "cedryl acetate": "77-54-3",
    "cimene para": "99-87-6",
    "citronellal": "106-23-0",
    "citronnellal": "106-23-0",
    "citronellol": "106-22-9",
    "citronnellol": "106-22-9",
    "citronelol": "106-22-9",
    "citronellool": "106-22-9",
    "citronellyl acetate": "150-84-5",
    "citronnellyl acetate": "150-84-5",
    "citronellyl formate": "105-85-1",
    "citronnellyl formate": "105-85-1",
    "cis 3 hexenol": "928-96-1",
    "cis-3 hexenol": "928-96-1",
    "hexenol cis-3": "928-96-1",
    "cis 3 hexenyl acetate": "3681-71-8",
    "cis 3 hexenyl benzoate": "25152-85-6",
    "cis 3 hexenyl salicylate": "65405-77-8",
    "cis-3 hexenyl salicylate": "65405-77-8",
    "cis jasmone": "488-10-8",
    "clonal": "106-72-9",
    "damascenone": "23696-85-7",
    "damascenone total": "23696-85-7",
    "damascone alpha": "43052-87-5",
    "alpha damascone": "43052-87-5",
    "damascone beta": "35044-57-6",
    "beta damascone": "35044-57-6",
    "damascone delta": "57378-72-0",
    "decalactone gamma": "706-14-9",
    "gamma decalactone": "706-14-9",
    "dihidromyrcenol": "18479-58-8",
    "dimethyl anthranylate": "85-74-5",
    "eucalyptol": "470-82-6",
    "eucaliptol": "470-82-6",
    "exaltex": "106-02-5",
    "estragol": "140-67-0",
    "estragone": "140-67-0",
    "floralozone": "67634-15-5",
    "frambinone": "5471-51-2",
    "heptalactone gamma": "105-21-5",
    "heptoate allyl": "142-19-8",
    "helvetolide": "141773-73-1",
    "hivernal neo": "300371-33-9",
    "hydoxycitronnellal": "107-75-5",
    "hidroxycitronellal": "107-75-5",
    "hidroxicitronelal": "107-75-5",
    "hydroxycitronnellal": "107-75-5",
    "indol": "120-72-9",
    "iso & super": "54464-57-2",
    "isobutyl quinoleine": "65442-31-1",
    "isobutylquinolein": "65442-31-1",
    "isobutylquinoleine": "65442-31-1",
    "isobornylcyclohexanol": "3407-42-9",
    "isobornyl cyclohexanol": "3407-42-9",
    "jasmonal h": "101-86-0",
    "jasmonal": "101365-33-7",
    "jessemal": "38285-49-3",
    "limonene": "5989-27-5",
    "macrolide": "106-02-5",
    "methyl anthranylate": "134-20-3",
    "methyl heptenone": "110-93-0",
    "methyl octin carbonate": "111-80-8",
    "methyl octine carbonate": "111-80-8",
    "methylnaphtylcetone": "93-08-3",
    "methyl naphtyl cetone": "93-08-3",
    "musc ambrette substitute": "83-66-9",
    "mus ambrette substitute": "83-66-9",
    "musc ketone": "81-14-1",
    "musk t": "105-95-3",
    "musc t": "105-95-3",
    "musc xylol": "81-15-2",
    "neo heliopan av": "5466-77-3",
    "nopyl acetate": "128-51-8",
    "norlimbanol": "70788-30-6",
    "oxyphenylon": "5471-51-2",
    "oxide rose l": "16409-43-1",
    "phenylacetic aldehyde": "122-78-1",
    "phenylpropyl alcohol": "122-97-4",
    "phenylpropylic alcohol": "122-97-4",
    "pinene alpha": "80-56-8",
    "pinene beta": "127-91-3",
    "rosalva": "13019-22-2",
    "rosacetol": "90-17-5",
    "rose oxyde": "16409-43-1",
    "rose oxyde d": "16409-43-1",
    "rose oxyde l": "16409-43-1",
    "salicilate hexenyl cis-3": "65405-77-8",
    "salicilate octyl": "118-60-5",
    "salicylate octyl": "118-60-5",
    "sandalor": "65113-99-5",
    "sandela": "3407-42-9",
    "sandranol": "28219-60-5",
    "terpinene gamma": "99-85-4",
    "terpenyl acetate": "80-26-2",
    "trimofix o": "144020-22-4",
    "veltol": "118-71-8",
    "veltol plus": "4940-11-8",
    "vertenex": "32210-23-4",
    "vetyver acetate": "117-98-6",
    "vetyveryle acetate": "117-98-6",
    "vetyveryl acetate": "117-98-6",
    "adoxal": "141-13-9",
    "amarocit": "67674-46-8",
    "amber extreme": "476332-65-7",
    "amberketal": "57345-19-4",
    "ambrinol": "41199-19-3",
    "andrane": "13567-39-0",
    "aurantiol": "89-43-0",
    "bourgeonal": "18127-01-0",
    "bourgeonnal": "18127-01-0",
    "brahmanol": "72089-08-8",
    "celeriax": "28231-03-0",
    "costus oliffac": "68917-29-3",
    "cuminic aldehyde": "122-03-2",
    "dimetol": "13254-34-7",
    "dynascone": "56973-85-4",
    "ethyl amyl ketone": "541-85-5",
    "hexalon": "79-78-7",
    "jacinth body": "1205-17-0",
    "koavone": "81786-73-4",
    "mandarinal": "540-18-1",
    "methylparacresol": "106-44-5",
    "methyl paracresol": "106-44-5",
    "methylpamplemousse": "67674-46-8",
    "mousse metra": "4707-47-5",
    "myrac aldehyde": "37677-14-8",
    "osyrane": "41890-92-0",
    "osyrol": "41890-92-0",
    "paracresol": "106-44-5",
    "paracresyl acetate": "140-39-6",
    "paracresyl phenylacetate": "101-94-0",
    "peryl aldehyde": "2111-75-3",
    "phenylethyl phenylacetate": "102-20-5",
    "phenylethyl dimethyl carbinol": "103-05-9",
    "prunolide": "104-61-0",
    "rhubafuran": "82461-14-1",
    "scatol": "83-34-1",
    "stemone": "22457-23-4",
    "syringa aldehyde": "104-87-0",
    "timberol": "70788-30-6",
    "vetylione": "79-89-0",
    "viridine": "101-48-4",
    # Existing natural-complex profiles.
    "ambreine pure": "NATURAL:68916-26-7",
    "animalys": "NATURAL:68916-26-7",
    "civet cnc": "NATURAL:68916-26-7",
    "civette synthetic": "NATURAL:68916-26-7",
    "civette synthetic / ambreine": "NATURAL:68916-26-7",
    "ambrarome": "NATURAL:8023-83-4",
    "ambrarome absolute": "NATURAL:8023-83-4",
    "ambrarome resinoid": "NATURAL:8023-83-4",
    "cardamome oil": "NATURAL:8000-66-6",
    "cardamome": "NATURAL:8000-66-6",
    "caraway": "99-49-0",
    "caryophyllene rect": "87-44-5",
    "cariophyllene rect": "87-44-5",
    "caryphyllene rect": "87-44-5",
    "cariophylene rect": "87-44-5",
    "cinnamon leaves": "97-53-0",
    "cinnamon leaves oil": "97-53-0",
    "cinnamon ceylon": "97-53-0",
    "cloves oil": "97-53-0",
    "clove bud": "97-53-0",
    "cumin oil": "122-03-2",
    "cumin eo": "122-03-2",
    "cumin seeds": "122-03-2",
    "cedarwood": "77-53-2",
    "cedarwood oil": "77-53-2",
    "cedarwood virginia": "77-53-2",
    "cedar virginia": "77-53-2",
    "clary sage oil": "115-95-7",
    "clary sage absolute oil": "115-95-7",
    "officinalis sage oil": "115-95-7",
    "armoise": "8023-88-9",
    "artemisia": "8023-88-9",
    "artemisia oil": "8023-88-9",
    "basilic": "8015-73-4",
    "basilic oil": "8015-73-4",
    "bigarade oil": "NATURAL:8014-17-3",
    "ciste oil": "8016-26-0",
    "ciste absolute": "8016-26-0",
    "citronella ceylon": "8000-29-1",
    "citronnella oil": "8000-29-1",
    "citronela ceylan": "8000-29-1",
    "copahu balsam": "8001-61-4",
    "coriander": "NATURAL:8008-52-4",
    "coriander oil": "NATURAL:8008-52-4",
    "gaiacwood": "NATURAL:8016-23-7",
    "galbanum synthetic r": "68912-13-0",
    "ginger co2": "8007-08-7",
    "guaiacwood": "NATURAL:8016-23-7",
    "gurjum balsam": "8030-55-5",
    "hysope oil": "84603-66-7",
    "immortelle absolute": "90131-24-1",
    "iris concrete": "90045-43-5",
    "irone": "79-69-6",
    "irone alpha": "79-69-6",
    "jasmin": "38285-49-3",
    "jasmin base": "38285-49-3",
    "laurus nobilis leaf": "8006-78-8",
    "liatrix absolute": "68606-83-7",
    "hyperessence liatrix": "68606-83-7",
    "lavander": "NATURAL:8000-28-0",
    "lavander oil": "NATURAL:8000-28-0",
    "lavander oil 40/42": "NATURAL:8000-28-0",
    "lavandin abryal": "NATURAL:8000-28-0",
    "lavandin abryal oil": "NATURAL:8000-28-0",
    "lavandin super oil": "NATURAL:8000-28-0",
    "mandarin": "NATURAL:8008-31-9",
    "mandarine": "NATURAL:8008-31-9",
    "mandarine oil": "NATURAL:8008-31-9",
    "mandarin vert": "NATURAL:8008-57-9",
    "mandarin verte": "NATURAL:8008-57-9",
    "mandarin italiana": "NATURAL:8008-57-9",
    "mandarine verte italie he": "NATURAL:8008-57-9",
    "mirceno": "123-35-3",
    "myrceno": "123-35-3",
    "myrte oil": "NATURAL:8000-25-7",
    "oakmoss": "90028-68-5",
    "oakmoss absolute": "90028-68-5",
    "absolute oakmoss": "90028-68-5",
    "orange brasil": "NATURAL:8008-57-9",
    "orange seet brasil oil": "NATURAL:8008-57-9",
    "orange sweet brasil oil": "NATURAL:8008-57-9",
    "patchouli decol": "5986-55-0",
    "patchouly": "5986-55-0",
    "patchouly dm": "5986-55-0",
    "patchouly dm oil": "5986-55-0",
    "patchouly oil": "5986-55-0",
    "petit grain": "NATURAL:8014-17-3",
    "petit grain oil": "NATURAL:8014-17-3",
    "petit grain paraguay": "NATURAL:8014-17-3",
    "pink pepper": "NATURAL:8000-25-7",
    "pepper oil": "NATURAL:8000-25-7",
    "pimento oil": "8006-39-1",
    "rose": "NATURAL:8007-01-0",
    "sandalwood oil": "NATURAL:8006-87-9",
    "vetyver oil": "NATURAL:8016-96-4",
    "vetiver haiti": "NATURAL:8016-96-4",
    "white thyme oil": "97-53-0",
    "white thyme": "97-53-0",
    "red thyme": "97-53-0",
    "red thyme oil": "97-53-0",
    "spearmint": "8008-79-5",
    "spearmint oil": "8008-79-5",
    "styrax oil": "8046-42-0",
    "taget oil": "8016-84-0",
    "tagets": "8016-84-0",
    "tarragon oil": "8016-88-4",
    "tonka beans absolute": "8024-04-2",
    "tonka beans resinoi": "8024-04-2",
    "violet leaves absolute": "8024-08-6",
    "ylang extra": "NATURAL:8006-81-3",
    "ylang extra oil": "NATURAL:8006-81-3",
    "ylang oil": "NATURAL:8006-81-3",
    "ylang substitute": "NATURAL:8006-81-3",
    "ylang ylang extra": "NATURAL:8006-81-3",
    "ylang ylang substitute": "NATURAL:8006-81-3",
    # Long-tail WiseMoor OCR/French/trade-name variants needed for row coverage.
    "citrone lool": "106-22-9",
    "citronelool": "106-22-9",
    "fleur d oranger base": "NATURAL:8014-17-3",
    "fixateur": "110-27-0",
    "fixateur 404": "110-27-0",
    "bornyl acetate": "76-49-3",
    "aspic": "NATURAL:8000-28-0",
    "aspic oil from spain": "NATURAL:8000-28-0",
    "blue camomile": "8015-92-7",
    "camomile blue": "8015-92-7",
    "roman camomile": "8015-92-7",
    "roman camomile oil": "8015-92-7",
    "camomile oil roman": "8015-92-7",
    "camomille romaine": "8015-92-7",
    "rose acetate": "141-12-8",
    "acetanisol crist": "100-06-1",
    "azarbre": "54464-57-2",
    "globalide": "34902-57-3",
    "magnolan": "27606-09-3",
    "salicylate benzyl": "118-58-1",
    "anise star": "8007-70-3",
    "formiate citronellyl": "105-85-1",
    "formiate de citronnellyle": "105-85-1",
    "formiate geranyl": "105-86-2",
    "nutmeg eo": "8008-45-5",
    "peppermint usp": "8006-90-4",
    "peppermint usa": "8006-90-4",
    "patchoily oil": "5986-55-0",
    "patchouli": "5986-55-0",
    "patchouly indo": "5986-55-0",
    "isobornyl acetate": "125-12-2",
    "ambrocenide": "211299-54-6",
    "ambrocenide 10": "211299-54-6",
    "cyclohexyl propionate allyl": "2705-87-5",
    "ciclohexyl propionate allyl": "2705-87-5",
    "allyl ciclohexyl propionate": "2705-87-5",
    "farenal": "141-13-9",
    "orange terpenes": "5989-27-5",
    "salicilate benzyl": "118-58-1",
    "salicilate hexyl": "6259-76-3",
    "salicilate amyl": "2050-08-0",
    "salicilate isoamyl": "87-20-7",
    "ally amyl glycolate": "67634-20-2",
    "allyl amyl glicolate": "67634-20-2",
    "aldehyde mandarin": "540-18-1",
    "lavandin grosso": "NATURAL:8000-28-0",
    "lavandin grosso oil": "NATURAL:8000-28-0",
    "lavandin": "NATURAL:8000-28-0",
    "lavandin abrialys oil": "NATURAL:8000-28-0",
    "acetate geranilo": "105-87-3",
    "acetyl methyl carbinol": "513-86-0",
    "butirate ethyl": "105-54-4",
    "caprilate ethyl": "106-32-1",
    "methyl butirate ethyl": "7452-79-1",
    "safranate ethyl": "35044-59-8",
    "2 methyl butyl acetate": "624-41-9",
    "elemi": "8023-89-0",
    "salvia sclarea french": "8016-63-5",
    "grapefruit": "NATURAL:8008-57-9",
    "acetate citronelyl": "150-84-5",
    "oxide caryophylene": "1139-30-6",
    "acid acetic": "64-19-7",
    "methyll heptenone": "110-93-0",
    "mircene": "123-35-3",
    "pheylethyl alcohol": "60-12-8",
    "phenylrthyl alcohol": "60-12-8",
    "phenylethyl alcohool": "60-12-8",
    "alcool phenylethylique": "60-12-8",
    "alcohol 39c": "60-12-8",
    "butirate d m b c": "10094-34-5",
    "dmbc butyrate": "10094-34-5",
    "fructalate": "72903-27-6",
    "monopropilenglicol": "57-55-6",
    "neobutenone alpha": "56973-85-4",
    "oriclone": "16587-71-6",
    "salicilate octilo": "118-60-5",
    "zenolide": "54982-83-1",
    "eucalyptus globulus 80": "470-82-6",
    "eucalyptus globulus oil": "470-82-6",
    "eucalyptus": "470-82-6",
    "diphenyl oxide": "101-84-8",
    "diphenyl oxyde": "101-84-8",
    "aldehyde c8": "124-13-0",
    "methyl dihydro jasmonate": "24851-98-7",
    "helvetolide pur": "141773-73-1",
    "canthoxal": "5462-06-6",
    "canthoxal iff": "5462-06-6",
    "cis 2 hxenyl acetate": "2497-18-9",
    "ethyl isovalerate": "108-64-5",
    "orange brazil": "NATURAL:8008-57-9",
    "ethyl aceto acetate": "141-97-9",
    "hidroxycitrone lool": "106-22-9",
    "hidroxycitronelool": "106-22-9",
    "hydroxycitonnellal": "107-75-5",
    "floropal": "5182-36-5",
    "d h m": "18479-58-8",
    "dihydrmyrcenol": "18479-58-8",
    "menta arvensis": "8006-90-4",
    "ciste labdanum sis absolute": "8016-26-0",
    "abs ciste labdanum": "8016-26-0",
    "cinnamyle acetate": "21040-45-9",
    "paramethyquinoleine": "491-35-0",
    "tetrahydroparamethylquinoleine": "635-46-1",
    "caprylic acid": "124-07-2",
    "linalyl propinate": "144-39-8",
    "birch tree oil": "8001-88-5",
    "valerianic acid": "109-52-4",
    "methylhexylcetone": "111-13-7",
    "carvyl acetate": "97-42-7",
    "phenylrthyl acetate": "103-45-7",
    "methyl phenyl acetate": "101-41-7",
    "methyl phenylacetate": "101-41-7",
    "phenylethyl isobutyrate": "103-48-0",
    "methyl phenylacetate": "101-41-7",
    "menthone": "10458-14-7",
    "isomenthone": "491-07-6",
    "rosalva iff": "13019-22-2",
    "ethyl citrate": "77-93-0",
    "benzyl butyrate": "103-37-7",
    "heliotrpin": "120-57-0",
    "heliotropin": "120-57-0",
    "gamma octalactone": "104-50-7",
    "c 12 l aldehyde 50": "112-54-9",
    # Formula-by-formula recovery for WiseMoor rows rejected by the coverage gate.
    "acetate de linalyle": "115-95-7",
    "acetate linalyle": "115-95-7",
    "acetate d isobornyle": "125-12-2",
    "ylang s": "NATURAL:8006-81-3",
    "geraniul egypt oil": "NATURAL:8007-01-0",
    "coriandre russe": "NATURAL:8008-52-4",
    "cypres biolande": "NATURAL:8014-17-3",
    "salicylate d amyle": "2050-08-0",
    "romarin": "NATURAL:8000-25-7",
    "thym rouge": "97-53-0",
    "cumin": "122-03-2",
    "black pepper": "87-44-5",
    "black pepper oil": "87-44-5",
    "cypress oil": "NATURAL:8014-17-3",
    "vetiver oil haiti": "NATURAL:8016-96-4",
    "vetiver oil bourbon": "NATURAL:8016-96-4",
    "cistus oil spain": "8016-26-0",
    "vetival": "117-98-6",
    "santalex": "NATURAL:8006-87-9",
    "tetrahtdromugol": "78-69-3",
    "tetrahydromugol": "78-69-3",
    "methyl pamplemousse giv": "67674-46-8",
    "veloutone": "65443-14-3",
    "veloutone firm": "65443-14-3",
    "bourgennal": "18127-01-0",
    "benzyl salicilate": "118-58-1",
    "magnolia oil lmr": "24851-98-7",
    "muscone": "541-91-3",
    "neo heliopan bb": "131-57-7",
    "abs jasmin maroc": "38285-49-3",
    "viridine padma": "101-48-4",
    "ethyl acetyl acetate": "141-97-9",
    "jasmin 1103": "38285-49-3",
    "dimethyl heptanol": "106-21-8",
    "myrthe oil": "NATURAL:8000-25-7",
    "lime distilled": "NATURAL:8008-57-9",
    "phenylethy alcohol": "60-12-8",
    "tagete oil": "8016-84-0",
    "marjolaine": "NATURAL:8000-25-7",
    "thyme oil": "97-53-0",
    "marjoliane dl": "NATURAL:8000-25-7",
    "orange": "NATURAL:8008-57-9",
    "vetyverol": "117-98-6",
    "verbena oil": "106-22-9",
    "clary sage absolute": "8016-63-5",
    "cedar leaf oil": "8007-20-3",
    "ylang absolute oil": "NATURAL:8006-81-3",
    "civet synth firm": "NATURAL:68916-26-7",
    "jasmin 231 type": "38285-49-3",
    "musc k": "81-14-1",
    "oakmoss absolute solute": "90028-68-5",
    "rose absolute type": "NATURAL:8007-01-0",
    "ambretolide": "7779-50-2",
    "cedarwood texas oil": "NATURAL:8006-87-9",
    "cedarwood oil texas": "NATURAL:8006-87-9",
    "dimethyl phenyl ethyl carbynol": "103-05-9",
    "dipropylen glycol": "25265-71-8",
    "isopropyl miristate": "110-27-0",
    "iso propyl myristate": "110-27-0",
    "sage sclarea oil": "8016-63-5",
    "vetyvert haiti oil": "NATURAL:8016-96-4",
    "ambreine": "NATURAL:68916-26-7",
    "camphre": "76-22-2",
    "cedarleaf oil": "8007-20-3",
    "copaiba balsam oil": "8001-61-4",
    "cuminaldehyde": "122-03-2",
    "labdanum oil": "8016-26-0",
    "tonka bean absolute": "8024-04-2",
    "cedre he bois atlas": "NATURAL:8006-87-9",
    "santal he inde": "NATURAL:8006-87-9",
    "acetate cedryle": "77-54-3",
    "vetyver ess java": "NATURAL:8016-96-4",
    "petitgrain he paraguay": "NATURAL:8014-17-3",
    "salicylate cis 3 hexenyle": "65405-77-8",
    "phenylethyl formate": "104-62-1",
    "amyl butyrate iso": "106-27-4",
    "ethyl propionate": "105-37-3",
    "boccoxim": "81782-77-6",
    "methyloctalactone": "39212-23-2",
    "clary sage absolute oil": "8016-63-5",
    "c 19 aldehyde": "104-67-6",
    "speatmint": "8008-79-5",
    "dihydrojasmone": "95-41-0",
    "taget absolute": "8016-84-0",
    "irisone alpha giv": "79-69-6",
    "rose bulgarian oil reconstitued": "NATURAL:8007-01-0",
    "ylang extra absolute": "NATURAL:8006-81-3",
    "carbitol": "111-90-0",
    "methyl cyclo pentenolene butyrate": "80-71-7",
    "sauge sclaree russe": "8016-63-5",
    "girofles clous": "97-53-0",
    "acetate de neryle": "141-12-8",
    "acetate de geranyle": "105-87-3",
    "acetate de citronnellyle": "150-84-5",
    "gurjum oil": "8030-55-5",
    "poivre": "87-44-5",
    "oxyde de rose d": "16409-43-1",
    "isobutyl isobutyrate": "97-85-8",
    "citronnelle java": "8000-29-1",
    "orange florida": "NATURAL:8008-57-9",
    "gamma terpinene": "99-85-4",
    "dilmrthyl anthranylate": "85-74-5",
    "thymol": "89-83-8",
    "alcool phenylethylique": "60-12-8",
    "geraniol": "106-24-1",
    "menthyl acetate": "89-48-5",
    "menthe bresil": "8006-90-4",
    "c 10 adehyde": "112-31-2",
    "citeronnellyl acetate": "150-84-5",
    "ethyl caproate": "123-66-0",
    "aldehyde c 7 e": "111-71-7",
    "methyl nonyl ketone": "112-12-9",
    "isopulegol purified": "89-79-2",
    "para cymene": "99-87-6",
    "ald c 11 undecylic": "112-44-7",
    "tangerine oil": "NATURAL:8008-57-9",
    "dioctyl adipate": "123-79-5",
    "dipentene": "5989-27-5",
    "lime terpenes": "5989-27-5",
    "c 9 aldehyde": "112-31-2",
    "vetyver terpenes": "NATURAL:8016-96-4",
    "animalys 1745": "NATURAL:68916-26-7",
    "butyric acid": "107-92-6",
    "amyris oil": "NATURAL:8006-87-9",
    "bouleau oil rectified": "8001-88-5",
    "beta pinene": "127-91-3",
    "gappefruit body": "NATURAL:8008-57-9",
    "amboxan henkel": "3738-00-9",
    "dynamone cas 8016 26 0": "8016-26-0",
    "dimyrcetol": "18479-58-8",
    "iso amyl butyrate": "106-27-4",
    "paratoluic aldehyde": "104-87-0",
    "hexyl formate": "629-33-4",
    "hexyl butyrate": "2639-63-6",
    "iso pulegol": "89-79-2",
    "citronnellyl isobutyrate": "97-89-2",
    "estragon oil": "8016-88-4",
    "diphenylmethane": "101-81-5",
    "c 9 alcohol": "143-08-8",
    "iso valerianic acid": "503-74-2",
    "hydratropic aldehyde dimethylacetal": "90-87-9",
    "hey absolute": "91-64-5",
    "galbex": "68912-13-0",
    "rosone bba": "16409-43-1",
    "cortex aldehyde": "68039-49-6",
    "c 11 alcool lenique": "112-42-5",
    "citeronnellyl acetate": "150-84-5",
    "methyl 2 pyrazinz": "109-08-0",
    "dibutyl sulfure iff": "544-40-1",
}


def _wisemoor_lookup_key(name: str) -> str:
    norm = unicodedata.normalize("NFKD", str(name).strip().lower())
    norm = "".join(ch for ch in norm if not unicodedata.combining(ch))
    norm = re.sub(r"[^a-z0-9&]+", " ", norm)
    return " ".join(norm.split())


WISEMOOR_NORMALIZED_OVERRIDES = {
    _wisemoor_lookup_key(raw): cas for raw, cas in WISEMOOR_NAME_OVERRIDES.items()
}


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser()
    parser.add_argument("--input", type=Path, default=Path("data/wisemoor_free_formulas.sanitized.jsonl"))
    parser.add_argument("--output", type=Path, default=Path("data/wisemoor_training_candidates.jsonl"))
    parser.add_argument("--audit-output", type=Path, default=Path("artifacts/wisemoor_training_candidate_audit.json"))
    parser.add_argument("--min-row-coverage", type=float, default=0.80)
    parser.add_argument("--min-amount-coverage", type=float, default=0.80)
    parser.add_argument("--min-resolved-components", type=int, default=3)
    return parser.parse_args()


def load_jsonl(path: Path) -> list[dict[str, Any]]:
    with path.open(encoding="utf-8") as handle:
        return [json.loads(line) for line in handle if line.strip()]


def normalize_formula(parts_by_cas: dict[str, dict[str, Any]]) -> list[dict[str, Any]]:
    total = sum(float(item["amount"]) for item in parts_by_cas.values())
    if total <= 0:
        return []
    formula = []
    for cas, item in sorted(parts_by_cas.items()):
        formula.append(
            {
                "cas": cas,
                "name": item["name"],
                "weight_fraction": float(item["amount"]) / total,
            }
        )
    return formula


def resolve_wisemoor_name(name: str, mapping: dict[str, str]) -> str | None:
    norm = " ".join(str(name).strip().lower().split())
    if norm in WISEMOOR_NAME_OVERRIDES:
        return WISEMOOR_NAME_OVERRIDES[norm]
    lookup = _wisemoor_lookup_key(name)
    if lookup in WISEMOOR_NORMALIZED_OVERRIDES:
        return WISEMOOR_NORMALIZED_OVERRIDES[lookup]
    compact = lookup.replace(" ", "")
    if compact in WISEMOOR_NORMALIZED_OVERRIDES:
        return WISEMOOR_NORMALIZED_OVERRIDES[compact]

    cas = map_ingredient(name, mapping)
    if cas:
        return cas

    aldehyde_match = re.match(r"^c\s+(\d+)(?:\s+([a-z]+))?\s+aldehyde$", norm)
    if aldehyde_match:
        chain, qualifier = aldehyde_match.groups()
        candidates = [
            f"aldehyde c-{chain} {qualifier}".strip(),
            f"aldehyde c-{chain}",
        ]
        for candidate in candidates:
            if candidate in WISEMOOR_NAME_OVERRIDES:
                return WISEMOOR_NAME_OVERRIDES[candidate]
            cas = map_ingredient(candidate, mapping)
            if cas:
                return cas

    return None


def convert_record(
    record: dict[str, Any],
    mapping: dict[str, str],
    *,
    min_row_coverage: float,
    min_amount_coverage: float,
    min_resolved_components: int,
) -> tuple[dict[str, Any] | None, dict[str, Any]]:
    raw_rows = record.get("formula", {}).get("formula_rows") or []
    rows = [
        row
        for row in raw_rows
        if re.search(r"[A-Za-z]", str(row.get("ingredient") or ""))
        and float(row.get("amount") or 0.0) > 0
    ]
    total_rows = len(rows)
    total_amount = sum(float(row.get("amount") or 0.0) for row in rows)

    parts_by_cas: dict[str, dict[str, Any]] = {}
    unresolved: list[dict[str, Any]] = []
    resolved_rows = 0
    resolved_amount = 0.0

    for row in rows:
        name = str(row.get("ingredient") or "").strip()
        amount = float(row.get("amount") or 0.0)
        if not name or amount <= 0:
            continue

        cas = row.get("cas") or resolve_wisemoor_name(name, mapping)
        if cas:
            resolved_rows += 1
            resolved_amount += amount
            existing = parts_by_cas.setdefault(cas, {"name": name, "amount": 0.0})
            existing["amount"] += amount
        else:
            unresolved.append({"ingredient": name, "amount": amount})

    row_coverage = resolved_rows / total_rows if total_rows else 0.0
    amount_coverage = resolved_amount / total_amount if total_amount else 0.0
    formula = normalize_formula(parts_by_cas)

    accepted = (
        row_coverage >= min_row_coverage
        and amount_coverage >= min_amount_coverage
        and len(formula) >= min_resolved_components
    )

    audit = {
        "product_id": record.get("product_id"),
        "name": record.get("name"),
        "total_rows": total_rows,
        "resolved_rows": resolved_rows,
        "row_coverage": row_coverage,
        "total_amount": total_amount,
        "resolved_amount": resolved_amount,
        "amount_coverage": amount_coverage,
        "resolved_components": len(formula),
        "accepted": accepted,
        "unresolved": unresolved,
    }

    if not accepted:
        return None, audit

    candidate = {
        "formula_id": f"WISEMOOR_{record.get('product_id')}",
        "source": "wisemoor_labtorium_free_sanitized",
        "source_url": record.get("source_url"),
        "name": record.get("name"),
        # Preserve source-observed formula text as provenance.  This is useful
        # for retrieval and later human/AI-assisted extraction, but is not an
        # assessor label and must not be promoted to one implicitly.
        "description": record.get("description"),
        "genre": "wisemoor_formula",
        "notes": record.get("notes"),
        "formula": formula,
        "metadata": {
            "product_id": record.get("product_id"),
            "slug": record.get("slug"),
            "row_coverage": row_coverage,
            "amount_coverage": amount_coverage,
            "unresolved_count": len(unresolved),
            "sanitized_source": str(Path("data/wisemoor_free_formulas.sanitized.jsonl")),
            "description_provenance": (
                "source_observed_wisemoor_product_text"
                if record.get("description") else None
            ),
        },
    }
    return candidate, audit


def write_jsonl(path: Path, records: list[dict[str, Any]]) -> None:
    path.parent.mkdir(parents=True, exist_ok=True)
    with path.open("w", encoding="utf-8") as handle:
        for record in records:
            handle.write(json.dumps(record, ensure_ascii=False, sort_keys=True) + "\n")


def main() -> None:
    args = parse_args()
    registry = AromaRegistry()
    mapping = build_tgsc_mapping(registry)

    candidates: list[dict[str, Any]] = []
    audits: list[dict[str, Any]] = []
    unresolved_counter: Counter[str] = Counter()

    for record in load_jsonl(args.input):
        candidate, audit = convert_record(
            record,
            mapping,
            min_row_coverage=args.min_row_coverage,
            min_amount_coverage=args.min_amount_coverage,
            min_resolved_components=args.min_resolved_components,
        )
        audits.append(audit)
        for item in audit["unresolved"]:
            unresolved_counter[item["ingredient"]] += 1
        if candidate:
            candidates.append(candidate)

    write_jsonl(args.output, candidates)

    by_reason = defaultdict(int)
    for audit in audits:
        if audit["accepted"]:
            by_reason["accepted"] += 1
        else:
            if audit["row_coverage"] < args.min_row_coverage:
                by_reason["low_row_coverage"] += 1
            if audit["amount_coverage"] < args.min_amount_coverage:
                by_reason["low_amount_coverage"] += 1
            if audit["resolved_components"] < args.min_resolved_components:
                by_reason["too_few_components"] += 1

    summary = {
        "source": str(args.input),
        "output": str(args.output),
        "records_total": len(audits),
        "records_accepted": len(candidates),
        "records_rejected": len(audits) - len(candidates),
        "thresholds": {
            "min_row_coverage": args.min_row_coverage,
            "min_amount_coverage": args.min_amount_coverage,
            "min_resolved_components": args.min_resolved_components,
        },
        "rejection_counts": dict(sorted(by_reason.items())),
        "top_unresolved": unresolved_counter.most_common(100),
        "records": audits,
    }
    args.audit_output.parent.mkdir(parents=True, exist_ok=True)
    args.audit_output.write_text(json.dumps(summary, ensure_ascii=False, indent=2) + "\n", encoding="utf-8")

    print(json.dumps({k: summary[k] for k in ["records_total", "records_accepted", "records_rejected"]}, indent=2))


if __name__ == "__main__":
    main()