Damselfly

Damselfly is a server-based photograph management application built for very large personal and professional photo libraries. Where most self-hosted gallery software focuses on browsing and sharing, Damselfly focuses on the management workflow that comes before sharing: indexing hundreds of thousands of images across deeply nested folders, building thumbnails in the background, applying IPTC keyword tags non-destructively, and finding the photo you actually need through fast multi-faceted search across keywords, folders, EXIF data, faces, and detected objects. Face recognition and object detection run entirely on the server with no third-party API dependency, so the AI-powered features that commercial photo services advertise are available without any of the privacy trade-offs.

Common Use Cases

Professional and serious-amateur photographers run Damselfly as the canonical index over a multi-decade RAW + JPEG archive, taking advantage of sub-second search across hundreds of thousands of images and using the basket workflow to pull selections into Lightroom, DigiKam, or Photoshop for editing. Wedding and portrait photographers tag and organise client deliverables, then expose curated subsets to assistants or to clients via per-folder access controls. Family historians and personal archivists run face recognition over decades of family photos, clustering relatives across childhood albums, holiday trips, and recent events into a single browse-by-person view that surfaces photos no manual tagging workflow would ever catch. Researchers and journalists curate evidence and image archives where IPTC keyword tagging matters for downstream syndication or rights management — Damselfly's non-destructive ExifTool-based writes preserve originals while updating metadata. Stock photographers and content creators add structured keywords to large submission batches before exporting them to stock-image platforms. Privacy-conscious users self-host Damselfly specifically to keep biometric face data, geotag history, and image content off the SaaS data graphs operated by major photo services, while still getting the same AI-powered organisation features.

Key Features

• Server-based indexing of very large photo libraries (500,000+ images supported with sub-second search)
• Local face detection and recognition with offline AI models — no third-party API dependency
• Object detection and image colour classification powered by YoloDotNet for content-based search
• Non-destructive EXIF/IPTC keyword tagging via ExifTool that preserves original JPEG encoding
• Background thumbnail generation, automatic library scanning, and incremental updates as files change on disk
• Multi-user accounts with admin and read-only roles, plus shared and per-user selection baskets
• Advanced search filters for date range, camera/lens make/model, file size, orientation, missing tags, and visual similarity
• Geolocation map view plotting geotagged photos on an interactive map
• Support for JPG, PNG, HEIC, TIFF, WebP, BMP, DNG, CR2, ORF, NEF, and other RAW formats
• Damselfly Desktop client (macOS, Windows, Linux) for syncing basket selections to a local folder for editing
• Watermark export and download workflows for social-media-ready exports and email delivery
• Folder reorganisation and trashcan-style deletion with a recoverable workflow
• `.nomedia` folder marker support for excluding subdirectories from scans

Why deploy Damselfly on Hostinger VPS

Running Damselfly on a Hostinger VPS gives photographers and personal archivists a private, vendor-independent home for image collections that need server-side AI features without leaking biometric data to a SaaS provider. Face vectors, geolocation history, and full-resolution image content are among the most sensitive personal data anyone manages, and self-hosting keeps every byte of it inside infrastructure you control. Dedicated CPU on a VPS handles the AI workload — face detection on every newly imported image, object recognition on the rest, EXIF parsing across the whole library — predictably, without competing with a desktop machine that comes and goes. Persistent volume storage keeps the SQLite metadata database, thumbnail cache, and the photo library itself durable across container restarts and host upgrades, and the photo volume can be replaced with a bind mount if you already have a structured library on the host. Combined with Traefik-based HTTPS routing on a clean branded hostname, the management UI is reachable from any browser over TLS, and the desktop client connects to the same hostname for sync-to-laptop editing workflows. Because the AI workloads benefit meaningfully from extra CPU and the .NET runtime plus thumbnail cache push memory usage to a couple of gigabytes, a mid-tier VPS plan is a better fit than the smallest plans — but the result is a serious photo-management server that scales to libraries hundreds of thousands of images deep.