Nazr SQLite v0.8.0

Why?

Pure speed. Nazr is built on Rust, which provides us with a significant advantage in raw performance, memory safety, and concurrency for I/O and CPU-heavy tasks, compared to a Node.js based backend, for example. It's light weight, which means faster deployments.

Customization. Most media management apps are opinionated, by design. Which organization strategy should the app follow: Metadata over folder structure? Should a delete only remove from the gallery, or the file system? We want to put these choices in your hands, offering sensible "set it, and forget it" preferences.

Superior Organization. Life is way too short to be spending on editing metatags, renaming countless files, and coming up with complicated folder structures. Why not let the software do the heavy lifting, and simply enjoy your photos.

Choice. For single-user simplicity, the SQLite backend steps away from a microservice architecture, to a self-contained binary. The ingestion pipeline is clear and modular for future extensibility. For even more scalability, there is also a PostgreSQL version.

Decoupled architecture. The backend deploys separately from the frontend, and is client agnostic. You can even create your own, and it will work with Nazr's backend.

Quirky - in all the good ways. A search box which accepts wildcards. Advanced filters to specifically search for WhatsApp or Google Pixel media. Extract audio from video. Ability to transcode older video formats such as MPG, 3GP, WMV, AVI, etc. Image burst capture of video.

Facial Recognition

Detect and mange individuals in your photo library. This feature uses InsightFace models (SCRFD for detection, ArcFace for recognition) via ONNX Runtime.

Requirements

• Models: The feature automatically downloads InsightFace models (~100MB total) on first run
• Memory: Models require ~200-300MB RAM when loaded
• Storage: Models are cached in data/models/ directory

Docker Images

All Docker images are available on Docker Hub at markraidc/nazr-backend-sqlite:

x86_64/AMD64 Images

With Facial Recognition:

• markraidc/nazr-backend-sqlite:0.8.0 (~867MB)
• markraidc/nazr-backend-sqlite:latest (~867MB)

Without Facial Recognition:

• markraidc/nazr-backend-sqlite:0.8.0-no-face (~829MB)
• markraidc/nazr-backend-sqlite:latest-no-face (~829MB)

ARM64 Images

With Facial Recognition:

• markraidc/nazr-backend-sqlite:0.8.0-arm64 (~214MB)
• markraidc/nazr-backend-sqlite:latest-arm64 (~214MB)

Without Facial Recognition:

• markraidc/nazr-backend-sqlite:0.8.0-no-face-arm64 (~214MB)
• markraidc/nazr-backend-sqlite:latest-no-face-arm64 (~214MB)

Pulling Images

# x86_64 with facial recognition
docker pull markraidc/nazr-backend-sqlite:latest

# x86_64 without facial recognition
docker pull markraidc/nazr-backend-sqlite:latest-no-face

# ARM64 with facial recognition
docker pull markraidc/nazr-backend-sqlite:latest-arm64

# ARM64 without facial recognition
docker pull markraidc/nazr-backend-sqlite:latest-no-face-arm64

---

Deployment Options

Nazr can be deployed with or without facial recognition support:

With Facial Recognition (Default)

Use the standard Docker Compose files:

• docker-compose.synology.yml
• docker-compose.ugreen.yml
• docker-compose.windows.yml

Benefits:

• ✅ Full facial recognition features
• ✅ Person detection and clustering
• ✅ Face tagging and organization

Requirements:

• ~200MB additional binary size
• ~200-300MB additional RAM for models
• ~100MB storage for model cache

Without Facial Recognition (Lightweight)

Use the -no-face Docker Compose files:

• docker-compose.synology-no-face.yml
• docker-compose.ugreen-no-face.yml
• docker-compose.windows-no-face.yml

Benefits:

• ✅ Smaller binary size (~200MB reduction)
• ✅ Lower memory usage (~200-300MB less RAM needed)
• ✅ Faster build times
• ✅ No model downloads required
• ✅ Ideal for resource-constrained devices

Trade-offs:

• ❌ No facial recognition features
• ❌ No person detection or tagging

Memory Requirements:

• With facial recognition: 1GB+ recommended
• Without facial recognition: 384MB-768MB sufficient

---

Windows Development

The docker-compose.windows.yml file automatically uses your Windows user's Pictures folder (%USERPROFILE%\Pictures).

Start the service:

docker compose -f docker-compose.windows.yml up --build

If you want to use a different photo directory, edit docker-compose.windows.yml and change the volume path:

volumes:
  - C:/path/to/your/photos:/photos:ro
  - ./nazr-data:/flash-data:rw

Lightweight Deployment (Without Facial Recognition)

To deploy without facial recognition support (saves ~200-300MB RAM):

docker compose -f docker-compose.windows-no-face.yml up --build

This version:

• Uses only 1.5GB RAM instead of 2GB
• Smaller binary size
• Faster build times
• No facial recognition features

Windows Native Build (EXE)

Build a native Windows executable for better file watcher support:

cargo build --release

Output: target\release\nazr-backend-sqlite.exe

Run with batch file (recommended):

run-nazr-windows.bat

The batch file automatically uses your Windows user's Pictures folder (%USERPROFILE%\Pictures).

Or run manually:

set FLASH_ROOT=%USERPROFILE%\Pictures
set FLASH_ROOT_HOST=%USERPROFILE%\Pictures
set FLASH_DATA=.\nazr-data
set FLASH_PORT=9161
set RUST_LOG=info

target\release\nazr-backend-sqlite.exe

Note: Native Windows builds use the image crate for thumbnails (slower than libvips but functional). File watcher works properly on Windows, detecting moves/renames in Explorer. Docker builds are unaffected.

---

Synology NAS Deployment

Prerequisites

Before deploying via Portainer or Docker Compose, create the required directories on your Synology NAS:

1. SSH into your Synology NAS or use File Station with admin privileges

2. Create the photo directory (if it doesn't exist):

   sudo mkdir -p /volume1/photos
   sudo chmod 755 /volume1/photos

3. Create the Nazr data directory:

   sudo mkdir -p /volume1/nazr
   sudo chmod 755 /volume1/nazr

4. Verify directories exist:

   ls -la /volume1/ | grep -E "photos|nazr"

Deployment via Portainer

1. Open Portainer in your browser
2. Navigate to Stacks → Add Stack
3. Name your stack (e.g., nazr)
4. Choose Git Repository or Upload method:
   • Git Repository: Point to your repository and select docker-compose.synology.yml
   • Upload: Copy the contents of docker-compose.synology.yml
5. Click Deploy the stack

Deployment via SSH/Command Line

docker compose -f docker-compose.synology.yml up -d --build

Configuration Notes

• Default photo path: /volume1/photos
• Data storage: /volume1/nazr (database, thumbnails, models)
• Port: 9161
• GPU Acceleration: Configured for Intel QSV (DS220+ with J4025 CPU)
• Memory: 1GB limit (suitable for DS220+ with 2GB RAM)

If your photos are in a different location, edit the docker-compose.synology.yml file and update the volume paths:

volumes:
  - /volume1/your-photos:/photos:rw
  - /volume1/nazr:/flash-data:rw

And update the environment variable:

environment:
  - FLASH_ROOT_HOST=/volume1/your-photos

Lightweight Deployment (Without Facial Recognition)

To deploy without facial recognition support (saves ~200-300MB RAM):

docker compose -f docker-compose.synology-no-face.yml up -d --build

This version:

• Uses only 768MB RAM instead of 1GB
• Smaller binary size
• Faster build times
• No facial recognition features

---

Ugreen NAS Deployment

Prerequisites

Before deploying via Portainer or Docker Compose, create the required directories on your Ugreen NAS:

1. SSH into your Ugreen NAS or use the file manager with admin privileges

2. Create the photo directory (if it doesn't exist):

   sudo mkdir -p /mnt/ugreen/photos
   sudo chmod 755 /mnt/ugreen/photos

   Note: Ugreen NAS paths may vary. Common alternatives:

   • /mnt/ugreen-nas/photos
   • /volume1/photos

   Check your actual mount points with: df -h or mount | grep mnt

3. Create the Nazr data directory:

   sudo mkdir -p /mnt/ugreen/nazr
   sudo chmod 755 /mnt/ugreen/nazr

4. Verify directories exist:

   ls -la /mnt/ugreen/ | grep -E "photos|nazr"

Deployment via Portainer

1. Open Portainer in your browser
2. Navigate to Stacks → Add Stack
3. Name your stack (e.g., nazr)
4. Choose Git Repository or Upload method:
   • Git Repository: Point to your repository and select docker-compose.ugreen.yml
   • Upload: Copy the contents of docker-compose.ugreen.yml
5. Important: Before deploying, verify the volume paths in the compose file match your Ugreen NAS structure
6. Click Deploy the stack

Deployment via SSH/Command Line

Important: First, verify and update the volume paths in docker-compose.ugreen.yml to match your Ugreen NAS:

# Check your actual mount points
df -h
mount | grep mnt

# Edit the compose file if needed
nano docker-compose.ugreen.yml

# Deploy
docker compose -f docker-compose.ugreen.yml up -d --build

Configuration Notes

• Default photo path: /mnt/ugreen/photos (update if different)
• Data storage: /mnt/ugreen/nazr (database, thumbnails, models)
• Port: 9161
• GPU Acceleration: Auto-detect (set to qsv if you have Intel CPU)
• Memory: 512MB limit (adjust based on your NAS specs)

Common Ugreen NAS path variations:

• /mnt/ugreen/photos
• /mnt/ugreen-nas/photos
• /volume1/photos

If your photos are in a different location, edit the docker-compose.ugreen.yml file:

volumes:
  - /your/actual/path/photos:/photos:rw
  - /your/actual/path/nazr:/flash-data:rw

And update the environment variable:

environment:
  - FLASH_ROOT_HOST=/your/actual/path/photos

Lightweight Deployment (Without Facial Recognition)

To deploy without facial recognition support (saves ~200-300MB RAM):

docker compose -f docker-compose.ugreen-no-face.yml up -d --build

This version:

• Uses only 384MB RAM instead of 512MB
• Smaller binary size
• Faster build times
• No facial recognition features

---

Service Access: After deployment, access Nazr at http://your-nas-ip:9161

Path Mapping (Docker/WSL)

When Nazr runs inside Docker/WSL, the database still contains the host's original file paths (for example C:\Users\you\Pictures). The backend now normalizes every file operation (rotation, delete-from-disk, etc.) by remapping that host prefix to the container mount. Configure these environment variables so the mapping is unambiguous:

• FLASH_ROOT – the path inside the container that is bind-mounted to your photos (defaults to /photos)
• FLASH_ROOT_HOST – the original host path that was scanned (e.g. C:\Users\you\Pictures)

Example for Windows Docker Compose:

environment:
  FLASH_ROOT: /photos
  FLASH_ROOT_HOST: C:\Users\you\Pictures
volumes:
  - C:/Users/you/Pictures:/photos:rw

With this mapping in place, Nazr can always resolve the real file even when running in a different environment.

Testing

Tests require a Linux environment and should be run using Docker (even on Windows/macOS):

# Run all tests
docker compose -f docker-compose.test.yml up --build

# Run specific test
docker compose -f docker-compose.test.yml run --rm test cargo test --test smoke --verbose

# Run without facial-recognition
docker compose -f docker-compose.test.yml run --rm test cargo test --no-default-features --verbose

Note: Tests are not designed to run natively on Windows/macOS. Use Docker for consistent, production-like testing. See TESTING.md for detailed documentation.

Notes

• Images/videos are not stored in SQLite, only metadata.
• Thumbnails and previews are saved under ${FLASH_DATA}/derived.
• SQLite is in WAL mode; DB file at ${FLASH_DATA}/db/nazr.db.