say hello

ClosetDelta

Wardrobe decision engine · Local MVP · Local MVP

A local wardrobe purchase decision app that scores whether a candidate item unlocks useful outfits against an existing closet, then saves the verdict, evidence, and feedback for deterministic scoring updates.

All work

What it is

ClosetDelta is for the moment before buying clothes, not for managing every outfit forever.

The product asks whether a new item is additive or redundant, then explains the verdict through outfit examples, compatibility signals, closest-overlap evidence, and saved decision history.

The loop it has to make believable

  1. 01

    Scoped the MVP to one decision instead of a broad closet-management app.

  2. 02

    Started with manual metadata and transparent rules so verdicts can be inspected and debugged.

How it’s built

The useful technical work is the scoring layer: compatibility, redundancy, seasonality, formality, and outfit unlocks.

The current runtime uses a JSON development store, not Prisma/Postgres, but it still models persistence concerns through snapshots, scoring-weight hashes, idempotent analysis creation, Zod API validation, and focused Vitest coverage.

Architecture

Inputs
Manual wardrobe and candidate metadata plus upload or source URL image input
Scoring
Category compatibility, color fit, season overlap, formality distance, redundancy, outfit unlocks
Output
Buy, Wait, or Skip verdict with score breakdown, outfit examples, overlap evidence, and explanation
Persistence
JSON-backed development store with candidate/wardrobe snapshots, scoring-weight hash idempotency, and feedback learning

What has to hold up

  • Persisted analyses with candidate and wardrobe snapshots so feedback remains meaningful after closet edits.
  • Added agree/disagree feedback that deterministically tunes pair-score weights from saved corrections.

How I would present it

I would present ClosetDelta as a product-thinking and explainability sample rather than a finished AI fashion app.

That framing is stronger because the current repo proves the decision logic, feedback loop, and local workflow, not computer vision accuracy, marketplace scraping, or production auth.

Proof

ClosetDelta is no longer just a scaffold: it has a local product loop, saved analyses, snapshots, and feedback learning, while still avoiding claims about CV, auth, or production persistence.

VerifyRepository, checks, stack, status
  1. 01

    Start with the project status and available evidence.

  2. 02

    Read the architecture and edge-case sections.

  3. 03

    Follow the documented proof path: README, app routes, dev store, scoring/feedback tests, Prisma schema.

Role
Product scope, scoring model, UI components, API shape
Scope
Local wardrobe-aware purchase decision workflow
Proof
README, app routes, dev store, scoring/feedback tests, Prisma schema
Stack
Next.js, React, TypeScript, Prisma, Tailwind CSS, Vitest
Verdicts
Buy/Wait/Skip
Scoring
Rule-based
Store
JSON dev store
Learning
Feedback tuned