Photarium is a self-hosted visual asset workflow for small studios, researchers, content teams, and self-hosters who want to ingest, curate, search, enrich, and publish image-heavy libraries without outsourcing storage or delivery to a generic DAM.
Photarium started as image management and now spans curation, ingestion, search, media handling, and adjacent client-page publishing while staying grounded in a simple self-hosted setup.
The current public story is local-first and operator-oriented. Photarium works best when you want your own internal visual library with practical tools around it rather than a glossy all-in-one hosted platform.
Keep working libraries of references, deliverables, generated assets, and variants in one place.
Build and search curated media sets with metadata, semantic helpers, and import-heavy workflows.
Handle web assets, campaign images, and reusable media without giving up direct CDN delivery.
Run the system locally or behind trusted internal access and keep the stack understandable.
The core product remains easy to evaluate: Cloudflare-backed asset management first, optional AI/search second, and more specialized workflows layered on top instead of replacing the basics.
The stable, daily-use surface for keeping a visual library tidy and usable.
Optional Redis- and API-backed workflows for discovery and machine-written metadata.
Bring media in from more than just a file picker when a library already exists somewhere else.
Push curated subsets outward without turning the internal app into a public-facing monolith.
Photarium is still designed for local or secured internal use. The external upload route does not ship with built-in authentication, so you should not expose the app directly to the public internet without adding auth middleware or trusted network protection.
These capabilities are real and already in the repo, but they belong in a clearly labeled advanced layer rather than the primary pitch for a first-time evaluator.
Scan pages for media, use browser-backed scroll mode when needed, and review queued candidates before upload.
Ingest video, inspect metadata, extract frames, and generate animated WebP artifacts alongside still images.
Use recursive ingest and source-specific helper scripts for local archives and adjacent social/media workflows.
Create explicit client-facing selects in Photarium and publish them to the adjacent public Cloudflare Worker app.
Detect embedded ComfyUI workflow data, store it as extras, and surface workflow context in asset detail views.
Backfill namespaces and embeddings, audit duplicates or broken assets, and maintain richer internal metadata stores.
The simplest evaluation path is still short: connect Cloudflare Images, add a few environment variables, and run it locally.
git clone https://github.com/bleeckerj/nfl-photarium.git and npm install.
Copy .env.example to .env.local and set Cloudflare account id, token, and account hash.
Start with npm run dev and open http://localhost:3000.
Add Redis for semantic discovery and OpenAI-backed helpers only if you need the AI/search layer.
Start with the core workflow first, then decide whether the AI/search and advanced workflow layers are worth enabling.
The repo is the source of truth. Use the Pages site as a concise overview, then jump into the docs that match the workflow you care about.
Start with the repo homepage for setup, requirements, and the high-level feature map.
Open Quick StartThe main operational guide for upload modes, search, embeddings, metadata, and maintenance scripts.
Open Features GuideReference for uploads, metadata updates, search routes, media workflows, and other API surfaces.
Open API ReferenceDetails for contract-based publishing into the adjacent public-facing Cloudflare Worker app.
Open Publishing Docs