Wearly AI Product Brief¶

Single source of truth for the Wearly AI product vision. This document drives all subsequent design, engineering, and roadmap decisions.

1. Product Name¶

The product name is Wearly AI.

2. Product Identity¶

Wearly AI is a mobile-first, clean luxury, professional personal styling agent. It helps busy women choose complete outfits using their real wardrobe, calendar, weather, location, body measurements, fit preferences, wear history, and favorite stores.

The product should later expand to support men, children, and family styling workflows, including helping parents coordinate outfits for their children.

3. Target User¶

The initial target user is women with busy schedules who need quick, context-aware outfit decisions for work, school, errands, appointments, travel, social events, and formal occasions.

Future versions should support:

men,
children,
family profiles, and
parent-managed child wardrobes.

4. Product Goal¶

The goal is to make outfit planning faster, more personal, and more context-aware.

Wearly AI behaves like an agent — it reasons end-to-end through:

calendar context,
weather,
wardrobe availability,
fit profile,
wear history, and
shopping preferences

before recommending an outfit.

5. UI Direction¶

The interface should be:

mobile-first,
clean luxury,
professional,
calm,
polished,
card-based,
easy to use, and
suitable for a premium lifestyle/productivity app.

It should not feel like a technical dashboard.

6. Realism Level¶

Treat this as a realistic, product-quality prototype, not a simple class demo.

Use real or realistic structured data. The product should feel like an early MVP that could eventually be launched.

7. Core Feature Areas¶

7.1 Calendar Integration¶

Wearly AI should support:

manual event entry,
Google Calendar integration or Google Calendar-ready architecture,
Apple Calendar / iPhone support through .ics upload in the prototype, and
future native iOS Calendar permission-based integration.

The agent should use:

event title,
time,
location,
description,
dress code (if available), and
event type

to infer occasion, formality, and styling needs.

7.2 Weather and Location¶

Wearly AI should support:

manual city selection,
optional current-location detection,
fallback to manual city entry if automatic location fails,
weather-based outfit reasoning, and
privacy notes explaining that location is used only for outfit planning.

Weather should influence clothing choices such as layers, shoes, fabrics, and outerwear.

7.3 Digital Wardrobe Builder¶

Users should be able to build their wardrobe by:

uploading or taking photos of clothing items,
removing the item background and replacing it with a white background,
adding items from product URLs from online stores, and
manually adding items.

Each wardrobe item should include metadata such as:

category,
color,
season,
formality,
fabric,
comfort,
modesty,
fit,
occasion suitability,
availability,
image,
source, and
ownership status.

7.4 Favorite Stores and Shopping¶

Users should be able to add favorite stores and shopping websites.

If the user has an event but does not own suitable pieces, Wearly AI should:

identify wardrobe gaps,
recommend missing pieces,
prioritize the user's preferred stores,
distinguish between owned items and suggested-to-buy items, and
allow saving suggestions to a wishlist or shopping list.

7.5 Fit and Body Profile¶

Users should be able to optionally enter measurements and fit preferences.

The profile may include:

height,
shoulder width,
bust / chest,
waist,
hips,
inseam,
preferred fit,
modesty preference,
comfort needs,
style goals,
areas to highlight, and
areas to balance.

The language must be body-positive. Never describe body features as flaws.

Use language such as:

highlight preferred features,
balance proportions,
support the user's preferred silhouette,
improve comfort, and
increase confidence.

The agent should use fit profile data in its reasoning and in its final explanation.

7.6 Wear History¶

Each wardrobe item should track:

how many times it has been worn,
last worn date,
event or occasion,
location,
outfit pairing,
user feedback, and
availability status.

Wearly AI should use wear history to:

avoid overusing the same item too often,
suggest underused items when appropriate, and
explain when wear history affects the recommendation.

7.7 Reject and Regenerate¶

Users should be able to reject:

a single item,
multiple items, or
the full outfit.

Rejection reasons may include:

wore it recently,
too formal,
too casual,
uncomfortable,
not suitable for weather,
does not fit well,
not modest enough,
do not like this color today,
unavailable, or
in laundry.

After rejection, the agent should regenerate a better recommendation and explain what changed and why.

7.8 Agentic Reasoning Trace¶

The app must clearly show that Wearly AI is an agent.

The reasoning trace should include:

Read event or user request.
Determine occasion and formality.
Check calendar.
Check weather and location.
Load wardrobe.
Load fit / body profile.
Apply styling rules.
Check color coordination.
Check wear history and availability.
Detect wardrobe gaps.
Recommend outfit.
Allow rejection and regeneration.

7.9 Before/After Demo¶

The project should include a clear before/after demo.

Before — A generic styling tool gives a simple, context-free outfit suggestion.
After — Wearly AI uses calendar, weather, wardrobe, fit profile, wear history, and favorite stores to create a personalized outfit with reasoning and alternatives.

8. Intelligent Book¶

Later, the repository should include a mini Intelligent Book.

The Intelligent Book should explain:

product vision,
user problem,
agent workflow,
styling knowledge base,
wardrobe intelligence,
fit profile logic,
calendar and weather context,
shopping gap logic,
before/after demo,
privacy and security, and
product roadmap.

The Intelligent Book supports the working app but does not replace it.

9. Knowledge Graph¶

Later, the repository should include a knowledge graph that models relationships between:

calendar events,
weather,
wardrobe items,
fit profile,
user preferences,
wear history,
outfit recommendations,
wardrobe gaps,
shopping suggestions, and
feedback.

The knowledge graph should help demonstrate that Wearly AI reasons through connected context rather than producing random outfit advice.

10. Agent Skills¶

Later, the repository should include a skills/ folder with a Wearly Styling Agent Skill.

The skill package should include:

SKILL.md,
occasion rules,
weather rules,
fit and silhouette rules,
wardrobe filtering rules,
color coordination rules,
wear history rules,
shopping gap rules, and
privacy guidelines.

11. Scientific and Expert Logic¶

Wearly AI's recommendation logic should not be based only on generic prompting.

It should combine:

evidence-informed fashion recommendation research,
color harmony principles,
body-positive fit and silhouette rules,
occasion and formality rules,
weather and location constraints,
wardrobe availability,
wear history,
user rejection reasons, and
user feedback.

11.1 References¶

The canonical references document is docs/evidence-and-references.md. It organizes source material into nine categories and tracks which categories rely on general practice vs. verified citations.

Source categories Wearly draws on:

Fashion recommender systems
Outfit compatibility
Color harmony
Body-shape-aware / fit-aware styling
Explainable recommendation systems
Human-centered AI / personalization
Wardrobe management / digital closets
Personalization and user feedback
Privacy and personal data

Verification status: as of this writing, zero specific citations have been verified. Every category currently relies on general practice and well-known frameworks (e.g., Itten color theory, GDPR, privacy-by-design). The references document lists 14 concrete pending verification items — specific searches a future contributor can complete.

The honesty contract:

Styling is not an exact science.
Wearly's logic is evidence-informed, design-informed, and user-preference-driven — never described as "correct" or "optimal."
Body-shape category labels (hourglass / pear / apple / etc.) are industry heuristics, not scientific taxonomy. Wearly treats them as user-declared proportion preferences only — never as classifications it computes.
Body-positive language is enforced as a runtime contract in code (fit_tool.check_reasoning_for_forbidden_language) and as a static contract across documentation (tests/test_documentation_language.py).
This brief, the Intelligent Book, and the Skill package never fabricate citations.

For per-rule sourcing, see the Source basis footer on every file in skills/wearly-styling-agent/. For the integrated narrative, see book/12-evidence-and-references.md.

12. SEIS 666 Alignment¶

This project is for SEIS 666 — Digital Transformation with AI.

Final project: Build Something That Reasons.
Track: Track B — Agentic AI System.

The project must demonstrate:

a working system,
a multi-step workflow,
at least two tools or data sources,
a documented decision process,
error handling,
a before/after comparison,
GitHub documentation, and
a 10-minute demo.

13. Product Roadmap¶

Phase	Scope
Phase 0	Fix current repo reliability.
Phase 1	Rebrand and improve the current demo.
Phase 2	Add product-grade mobile-first UI and user profile.
Phase 3	Add wardrobe builder, fit profile, wear history, and reject/regenerate.
Phase 4	Add shopping, wishlist, and favorite stores.
Phase 5	Add Intelligent Book, Knowledge Graph, and Skills.
Phase 6	Add production roadmap for real Google Calendar, Apple Calendar, image processing, store integrations, mobile app, privacy, and family profiles.

This document is the single source of truth for the Wearly AI product vision. All design, engineering, and roadmap decisions should align with this brief.