Each key is a theme (your keyword bucket); each value is either a boolean query or a plain list of terms. Queries support OR, AND, NOT, NEAR/3, parentheses, "quoted phrases", */? wildcards and #hashtags, and are matched against the words of each search term - so (vanilla OR caramel) NEAR/3 (perfume OR candle) NOT (recipe) behaves as written. Term lists keep the older forgiving substring matching (whole words for terms of 4 characters or fewer).

These options keep bare ingredient searches inside a theme. The first also accepts a product word, so "sandalwood candle" counts as Woody even though the definition only lists scent, fragrance, perfume, aroma, smell, bouquet and note. The second also accepts a search that is only notes and filler words, so "sandalwood" counts too - while "vanilla extract" and "rose garden" stay out. They only matter for queries written as term near context; plain term lists are unaffected.

Brands and formats are optional. Each key is a brand or a product type, and the value is a list of aliases matched against keywords the same way as terms. The bundled lists are a fragrance starter set; replace them with your own. Signature product names are included as aliases, so "santal 33" counts as Le Labo. Product types classify the format being searched for (candle, perfume, air freshener, wax melt).

When on, an unmatched search is attributed to a brand if one of its aliases is within one edit of a search word, and only when exactly one brand matches. Use it for messy exports; leave it off for clean data to avoid false positives.