Your closet, but with an offline AI stylist

What it does Wardrowbe is a self-hosted web app for cataloging clothes and planning outfits. You upload photos, a vision model auto-tags items by color, pattern, and style, and a text model recommends combinations based on weather, occasion, and your wear history. It tracks wash status, logs daily outfits, and serves analytics on what you actually wear.

The interesting bit The AI backend is deliberately vendor-agnostic: it talks to any OpenAI-compatible endpoint, so you can run entirely offline with Ollama on a Raspberry Pi or point it to OpenAI. You can also disable internal AI completely and use the app as a plain wardrobe database, deferring tagging and suggestions to an external agent.

Key highlights

• Fully self-hosted stack with pre-built multi-arch Docker images and Kubernetes manifests.
• Pluggable AI backend: swap between local Ollama, OpenAI, or LocalAI via environment variables.
• Optional AI master switch lets the backend boot without any LLM configured.
• Family support with per-member wardrobes, wear history, and outfit ratings.
• Scheduled daily outfit nudges through ntfy, Mattermost, or email.

Caveats

• Google Play support is listed as “Coming Soon”; mobile is currently iOS-only via the App Store.
• Default local login uses dev credentials with no password, so production deployments need proper auth configuration.

Verdict Worth a look if you want a privacy-first wardrobe manager and don’t mind self-hosting a Next.js/FastAPI stack. Not for anyone seeking a zero-setup SaaS or immediate Android access.