ClozeHive

Wardrobe intelligence, powered by FANI

A digital closet that thinks — outfit building, AI styling, travel packing, wear analytics, and RAG-grounded insights built on top of your own items and history.

Launch the app →  •  Marketing site  •  Repository

---

Table of contents

1. Overview
2. Meet FANI
3. Feature highlights
4. Architecture
5. Core workflows
6. Service responsibilities
7. Domain map (api-gateway)
8. Technology stack
9. Quick start (Docker)
10. Local setup without Docker
11. Configuration
12. API surface (v1)
13. Testing & quality
14. Operations & health
15. Frontend routes
16. Documentation

---

Overview

ClozeHive is a wardrobe intelligence platform. Users build a digital closet, compose and save outfits, get AI-driven styling and packing help, and track how they actually wear their clothes over time. Every recommendation is grounded in the user's real items and history through a retrieval-augmented (RAG) pipeline backed by vector search.

The system is a domain-driven architecture:

• a React + Vite + TypeScript single-page app,
• a FastAPI api-gateway that owns the full product API — identity, wardrobe, intelligence, travel, and platform domains — runs the image-analysis / background-removal vision pipeline in-process, and runs all FANI AI (chat, outfit, packing) in-process against OpenAI/Gemini with a RAG + grounding pipeline,
• PostgreSQL + pgvector as the single source of truth and embedding store, Redis for cache and session/OAuth state, and
• nginx as the single browser entrypoint in the local Docker topology.

Full environment, migration, GCP, and deployment detail lives in docs/ENVIRONMENT.md, docs/SERVICES.md, and docs/RUNBOOK.md. AI agents should read AGENTS.md for the operational contract.

---

Meet FANI

FANI — Fashion AI Nurturing Individuality — is ClozeHive's built-in AI stylist. FANI powers outfit recommendations, the daily Tip of the Day, wardrobe-gap detection, conversational styling, travel packing, and Shop with FANI purchase advice.

The primary stylist experience is the full-screen chat at /ai-stylist (persisted sessions, structured outfit cards, closet context). The dashboard surfaces FANI nudges and Tip of the Day.

Outfit suggestions carry a 5-tier FANI rating — Needs Improvement → Average → Good → Excellent → Best Fit — each with a star count, a color badge, and a short stylist note.

---

Feature highlights

Area	What you get
Home	Daily FANI Tip of the Day (day-of-week rotation), Style at a Glance (Most Worn / Never Worn / Closet Gaps), quick-action grid, and Outfit of the Day with a 5-tier FANI rating.
Style onboarding	Guided style-profile flow (/onboarding/style-profile) that feeds FANI personalization and RAG context.
Closet	Full item CRUD with rich metadata, image storage (local disk or GCS), preview/confirm upload flows, similar-item search, availability flags, and AI background removal.
Closet match	Find items in your closet that pair with a reference piece (/closet-match).
Smart ingest	Upload a photo (incl. HEIC) → detection → background removal → structured item metadata, all in-process in the gateway. Bulk smart-ingest sessions for multi-photo review.
Outfits	Outfit builder, saved looks, outfit history (feeds RAG), and weekly planner with weather-aware day plans.
FANI AI Stylist	Structured stylist chat at /ai-stylist with SSE streaming, persisted history, and handoff from dashboard nudges.
Travel	Trips, weather-aware packing lists, packing-memory APIs, occasion planning, and FANI/agent packing tools.
Shop with FANI	Upload an in-store photo or paste a product URL → closet match, buy/skip verdict, outfit ideas, and optional add-to-closet (SSRF-guarded URL fetch).
Intelligence (RAG)	Fashion knowledge, closet similarity, purchase-gap detection, and a unified /rag/* surface backed by pgvector embeddings.
Analytics	Wear events (append-only), cost-per-wear, utilization, forgotten gems, and versatility scoring.
Profile & Settings	Display name, location & weather, FANI preferences, avatar, and privacy; navbar dropdown for Saved Outfits, Closet Insights, and Wardrobe Gaps.
Real-time	WebSocket notifications at /api/v1/ws (toast UI in the app shell).

---

Architecture

ClozeHive runs in two topologies that share one codebase. The gateway is the single owner of the wardrobe and intelligence domains and runs the vision pipeline in-process in both.

Production (Render) — canonical

The frontend calls the gateway directly (VITE_API_URL). There is no nginx; the gateway serves the entire /api/v1 surface with SERVE_MIGRATED_DOMAIN_ROUTES=true and runs all AI in-process.

This is the beginning of the README. Read the rest at the original. Read the full README ↗