Self-hosted AI photo manager. Your Pictures. Fast. Easy. Simple.
Warning
Eyeris is designed for use on a trusted local network only. It has no brute-force protection on the login endpoint, no multi-user access control, and image URLs embed session tokens that appear in server logs. Do not expose it directly to the internet. If you need remote access, use a VPN or put it behind Cloudflare Access — not a plain public tunnel.
| Who it's for | Method | Guide |
|---|---|---|
| Unraid users | Manual install via template — see the wiki | Unraid → |
| Everyone else (Linux / Windows / macOS with Docker) | Run ./docker-start.sh |
Docker → |
| Developers | Run ./start.sh — Python + Node directly, no Docker |
Dev → |
Not sure? Use the Docker path. It works on any OS and handles everything automatically.
# 1. Clone the repo
git clone https://github.com/vonhex/Eyeris.git
cd Eyeris
# 2. Run the start script — it will create .env and tell you what to fill in
chmod +x docker-start.sh
./docker-start.sh
# 3. Edit .env with your NAS credentials (SMB_HOST, SMB_USERNAME, etc.)
nano .env
# 4. Run again — pulls the image and starts the container
./docker-start.shOpen http://localhost:8000 — default username is eyeris, password is set on first run.
- Infinite-scroll gallery with thumbnails, favorites, and multi-select bulk actions
- Filter by tag, category, album, folder, camera, location, date range, GPS, or quality issue
- Sort by date taken, date added, filename, or random
- Video support (MP4, MKV, MOV, AVI, WebM and more) with thumbnail preview
- Untagged view shows both images and videos with no tags yet
- Desktop: mouse-wheel zoom (up to 8×), click-drag to pan, double-click to toggle zoom, keyboard ← → to navigate
- Mobile: full-screen view, pinch-to-zoom, horizontal swipe to navigate between images, swipe-up bottom sheet to reveal metadata and editing controls
- Scroll-to-top button appears site-wide when scrolling long pages
- SigLIP zero-shot tag classification against ~300 candidates — runs locally on GPU during scan
- YOLOv8 + FaceNet face detection and bounding-box metadata per image
- XMP sidecar ingestion: tags and descriptions written by external tools are read automatically during scan
- Connect to a self-hosted A-Eye instance from the Settings page
- Send untagged images and videos to A-Eye for AI analysis with one click from the Dashboard
- Video thumbnails are sent automatically when a video has no tags
- Tags and descriptions written by A-Eye are auto-imported every 5 minutes — no manual re-import needed
- Supports HTTP Basic Auth for protected A-Eye instances
- A-Eye URL, username, and password persist across container updates
- Date taken, GPS location with OpenStreetMap link, reverse-geocoded city/country
- Camera make/model, lens model, aperture, shutter speed, ISO, focal length
- Tag editor per image + bulk tag editor across selections
- Category and album assignment
- Duplicate detection via perceptual hash
- Two-phase async scan pipeline with live progress bar
- File-watcher for near-real-time new image detection
- Settings page with runtime
.envediting — all settings survive container recreation - Dashboard shows accurate untagged counts for both images and videos
PolyForm Noncommercial License 1.0.0 — free to use and modify, not for commercial use or resale. For commercial licensing contact lucashjantzen@gmail.com.