Atlas – Your photos and files on your own server

Atlas is a self-hosted photo library and drive for a single home server, with one iOS app in front of it. The backend is Rust end to end; it runs on your own hardware and is reached over a Tailscale tailnet.

---

Features

• Photo library timeline of every photo and video, albums, people, places, favorites, archive, a locked album and a trash
• Search by content "dog in the snow" finds it: one embedding model for photos, videos and text, scanned exactly in memory
• Faces detection and recognition group photos into people you can name and merge
• Backup the iPhone uploads originals, deduplicated by content hash, and can clear backed-up items off the phone
• Share links albums or a few photos as a link (optional password), served from your own Cloudflare account so links work while the server sleeps, and gone after 7 days at the latest
• Drive folders over content-addressed files, with previews, full-text search and a trash
• One app Fotos, Alben, Dateien and Einstellungen, with live server status, power control and a terminal
• Built to feel instant precomputed and compressed timeline, immutable media URLs, streaming uploads, video renditions on the GPU
• Wake-on-LAN CLI atlas boot wakes the server, atlas deploy installs the services, atlas build and atlas dev build other projects on it
• Tailnet-first security one port, firewalled to loopback and the tailnet, one bearer token on every route

---

Quick start

# Mac: install the CLI, then configure your machine values (see docs/SETUP.md)
cargo install --path crates/cli
mkdir -p ~/.config/atlas && $EDITOR ~/.config/atlas/env

atlas boot        # wake the server (Wake-on-LAN)

# Server: database, models, configuration
cd ~/atlas/db && cp .env.example .env && docker compose up -d
~/atlas/scripts/atlas/models.sh
sudo install -m600 ~/atlas/scripts/atlas/atlas.env.example /etc/atlas/atlas.env

# Mac again
atlas deploy      # build + install atlas-server and atlas-ml
atlas connect     # the link that connects the app

---

Documentation

• Overview: repository layout, architecture, security
• Setup: from-scratch machine setup (Ubuntu, Tailscale, Wake-on-LAN, GPU, models, the app)
• Components: server · ml · cli · app · db · builder · scripts

---

License

MIT License - View License
Model weights downloaded at setup (Qwen3-VL-Embedding, InsightFace buffalo_l with its non-commercial research license) keep their own licenses.

---

Support

• Report bugs
• luka@lukaloehr.com

---

Developed by Luka Löhr