Occasion rules¶

Mapping free-text occasions to canonical tags; required piece-types per occasion. Implemented in styling_agent.py.

R1 — Tag mapping¶

Free text	Canonical tag
`work`, `business`, `meeting`	`work`
`gym`, `workout`, `yoga`, `running`	`gym`
`dinner`, `restaurant`, `date`	`dinner`
`formal_event`, `gala`, `formal`	`formal`
`casual`, `weekend`, `brunch`, `everyday`	`casual`

Unknown free-text → casual (safe default).

R2 — Required piece-types¶

Occasion	Required pieces
`work`	`top` + `bottom`
`dinner`	`top` + `bottom` (unless a dress is preferred)
`gym`	two `activewear` pieces
`formal`	`dress`
`casual`	`top` + `bottom`

Step 6 checks the built outfit against this list. Missing required pieces become gaps.

R3 — Dress-vs-separates branching¶

If occasion ∈ {formal, dinner} AND formality ≥ smart_casual: - Prefer a dress from the wardrobe pool. - For formality = formal, sort dresses to prefer formality = formal first.

Otherwise, build top + bottom (or two activewear pieces for gym).

R4 — Accessory cap¶

Add up to two accessories from the pool. Never more — over-accessorizing is the most common visual misstep in algorithmically-built outfits.

Source basis¶

Primary category: Outfit compatibility (docs/evidence-and-references.md §3.2). Secondary: Fashion recommender systems (§3.1). Current basis: Industry styling heuristics. The five-occasion taxonomy (work / dinner / gym / formal / casual) is broadly recognized in fashion-product UX. No specific peer-reviewed paper cited yet — see verification status at docs/evidence-and-references.md §7.