The AI's clothing-type vocabulary is a hardcoded set (VALID_TYPES in ai_service.py plus the clothing_analysis prompt), around 30 generic types. Anything the model returns outside that set is coerced to "unknown", so there's no way to extend it without editing source.
The list skews Western/masculine and misses many common garments: kimono, camisole, playsuit, bodysuit, gown, kaftan, tunic, culottes, dungarees, bralette, leggings, and more. In practice blouses get tagged "shirt", trousers "pants", and femme wardrobes feel poorly served.
Request
Let users add and edit the category list (and ideally the colour / pattern / style vocabularies) in settings, with the tagging prompt and validation reading from that config rather than a hardcoded set.
Workaround today
An item's type is free-form on update so it can be hand-edited, but new uploads still use the fixed list.
The AI's clothing-type vocabulary is a hardcoded set (
VALID_TYPESinai_service.pyplus theclothing_analysisprompt), around 30 generic types. Anything the model returns outside that set is coerced to "unknown", so there's no way to extend it without editing source.The list skews Western/masculine and misses many common garments: kimono, camisole, playsuit, bodysuit, gown, kaftan, tunic, culottes, dungarees, bralette, leggings, and more. In practice blouses get tagged "shirt", trousers "pants", and femme wardrobes feel poorly served.
Request
Let users add and edit the category list (and ideally the colour / pattern / style vocabularies) in settings, with the tagging prompt and validation reading from that config rather than a hardcoded set.
Workaround today
An item's
typeis free-form on update so it can be hand-edited, but new uploads still use the fixed list.