Skip to content

11 · Product roadmap

The app is the prototype state; this chapter explains where the agent goes next and why each phase is ordered the way it is.

Where Wearly is going. Mirrors the roadmap in Wearly_Product_Brief.md §13 but explains each phase in product-engineering terms.

Status today

Phase Title State
0 Foundation reliability ✅ Shipped
1 Rebrand + demo polish ✅ Shipped
2 Mobile-first UI shell + product screens ✅ Shipped
4 (partial) Reject & regenerate ✅ Shipped
6 Intelligent Book + Knowledge Graph + Agent Skill package 🟡 In progress (this book is part of it)
7 (partial) Smoke tests + architecture doc ✅ Shipped
3 Data model split + fit profile wiring Not started
4 (remaining) Wardrobe builder + wear history rotation Not started
5 Shopping + wishlist + favorite stores Not started
7 (remaining) CI + final demo polish In progress

Near-term (before final demo)

Deploy and capture hero

  • Push the branch to GitHub.
  • Deploy to Streamlit Community Cloud (one click after push).
  • Capture a 60–90 second screen GIF of the full flow: home → tap CTA → outfit + reasoning → reject an item → regenerate → "what changed" banner.
  • Update the README hero block with the live URL and the embedded GIF.

First chapter of Intelligent Book ✅

Done in this batch — chapters 01 through 11 of this book, plus the knowledge graph and skill package.

CI workflow

GitHub Actions running python -m unittest discover tests on every push. A green badge in the README.

Medium-term (after final demo)

Phase 3 — Data model split

Move owner out of wardrobe.json into:

  • data/user_profile.json — name, locale, city, privacy preferences.
  • data/fit_profile.json — all the optional measurement and preference fields.

Add fit_tool.py to wire the fit profile into Step 5 of the agent.

Phase 4 (remaining) — Wardrobe builder

  • Add-by-photo screen with a Streamlit file uploader.
  • White-background composite (prototype simulation; real rembg behind an optional install).
  • Add-by-URL flow with manual field confirmation (real per-store scraping is future production).
  • Edit / delete on any wardrobe item.
  • data/wear_history.json driving rotation.
  • Per-item reject popovers on the outfit card (the multiselect form ships in Phase 2; per-item is a polish pass).

Phase 5 — Shopping & wishlist

  • data/stores.json — favorite stores.
  • data/wishlist.json — saved suggestions.
  • shopping_tool.suggest(gap, stores) returning store-prioritized cards.
  • Wishlist screen.

Long-term (production)

Native mobile app

Streamlit is the prototype surface. Real users will live in a native iOS app (Swift / SwiftUI) with the agent running server-side. The Streamlit prototype's API contract becomes the server's API contract.

Real integrations

  • Google Calendar (OAuth).
  • Apple Calendar (CalDAV or native iOS Calendar permission).
  • Geolocation (opt-in, coarse).
  • Affiliate-compliant shopping catalogs.

Family workflows

  • Parent-managed child profiles.
  • Coordinated family outfits ("dress us alike for the family photo").
  • School uniform support.
  • Stricter privacy controls on children's data.

Aesthetic learning

A bounded ML layer: a small model fine-tuned per user that learns aesthetic preferences from accept/reject history. Outputs feed back into the rule layer, never replace it. The reasoning trail stays explainable.

What's deliberately not on the roadmap

  • Infinite influencer feed. Wearly is decision support, not a scroll.
  • Social sharing of outfits. The user's wardrobe is their own.
  • Body image guidance. Wearly doesn't tell users how their bodies should look.
  • One-click checkout to retailers. Affiliate commerce is incompatible with neutral styling advice.

How to read this roadmap

The phases are loose, not contractual. The roadmap exists to make trade-offs visible: each new feature has a phase number so you can ask "why is this here and not there?" and get a structural answer instead of a politics answer.

Where to verify

  • The canonical roadmap: Wearly_Product_Brief.md §13.
  • The current state: git log --oneline + the Status today table above.

Key terms

  • Skill package — the skills/wearly-styling-agent/ rule pack (8 rule files + SKILL.md) (glossary).
  • Result-dict contract — the stable boundary that lets the UI and agent evolve independently (glossary).
  • Wishlist — the locally-saved taste signal that biases future recommendations (glossary).

Self-check

  1. (Remember) Name one feature shipped today and one feature on the roadmap.
  2. (Understand) Why is the result-dict contract called out as a stability anchor in the roadmap?
  3. (Apply) Pick a roadmap item. Identify one rule pack and one chapter that would need to change to land it.

Definitions live in the Glossary. Self-check questions follow Bloom's taxonomy progression (Remember → Understand → Apply → Analyze).