Stationary

Multi-Platform Media Asset Management & Integration Platform

A high-performance, self-hosted media asset manager (MAM) designed for creative professionals and content managers. Centralize, archive, organize, and search multi-platform media content ingested from external sync agents with interactive playback and hybrid AI search.

Key Features • User Interface • Monorepo Workspace • Quick Start • Documentation

简体中文文档

---

✨ Key Features

• 🎨 Card-on-Canvas Design System: Built on Apple-inspired spatial depth layers and Linear-class desktop refinement with clean HSL palettes.
• 📸 Native Live Photo & Interactive Media: Support for Live Photo interactive playback and browser-native HEIC/HEIF decoding.
• 🎬 Unified DASH Streaming: Audio/video split streams from multi-track media sources are parsed on-the-fly via dynamic sidx SegmentBase extraction and streamed smoothly with dash.js.
• 🏷️ Multi-Tenant Library Management: Physical asset isolation per Library, fine-grained RBAC (VIEWER, EDITOR, ADMIN), and reconstructive tag normalization with alias mapping.
• 🔍 Hybrid AI Search & Enrichment: Reciprocal Rank Fusion (RRF) combining SQL trigram text search, Gemini text embeddings (text-embedding-004), and visual similarity vectors (multimodal-embedding-004).
• ⚡ Durable Task Queue Engine: In-house PostgreSQL-backed task queue with atomic SKIP LOCKED lease locking, exponential backoff retries, and automated JobSweeper cleanup.

---

🎨 User Interface & Views

1. Board View (Posts)

Presents content grouped by Posts. Each card displays author details, publication timestamps, tags, and a thumbnail of the primary media asset.

2. Inspector Drawer (Post Details)

Slides out from the right when selecting any post. Features a Swiper-powered interactive media carousel alongside complete post metadata inspectors.

3. Media Grid (All Pins)

Directly displays individual Media assets with two flexible layout modes:

• Flat Mode: Displays all images and videos as independent cards for granular searching and asset inspection.
• Stacked Mode: Groups media belonging to the same post into a single card with stack badges (+N).

---

🛠️ Tech Stack & Monorepo

Stationary is structured as a Bun Workspace Monorepo:

stationary/
├── apps/
│   ├── server/       # Bun + Hono API backend, Drizzle ORM, DB Task Queue Engine
│   ├── web/          # Nuxt 4 + Vue 3 SPA/SSR, Tailwind CSS v4, Swiper, Plyr
│   ├── flutter/      # Flutter multi-platform client application
│   └── apple/        # Native iOS 18+/iPadOS/macOS app (SwiftUI, Liquid Glass, Swift 6)
└── docs/             # Technical specifications & architecture blueprints

Component	Stack	Responsibilities
apps/server	Bun, Hono, Drizzle ORM, PostgreSQL, Redis, @ai-sdk/google	REST APIs, authentication (Better Auth), task execution engine, S3 asset handling
apps/web	Nuxt 4, Vue 3, Pinia, Vue Query, Tailwind CSS v4, dash.js	Web desktop client, Card-on-Canvas layout, DASH video player, HEIC renderer
apps/flutter	Flutter, Dart, Chewie, PhotoView	Mobile and desktop cross-platform application
apps/apple	SwiftUI, Swift 6, AVKit, Liquid Glass Theme	Native iOS/iPadOS/macOS multiplatform app with Liquid Glass UI, Post List & Detail

---

🚀 Quick Start

Prerequisites

• Bun: v1.3.0 or higher
• PostgreSQL: v15 or higher (with pgvector & pg_trgm extensions enabled)
• Redis: For KV caching and rate limiting

1. Install Dependencies

bun install

2. Environment Configuration

Set up environment configuration files:

• Backend: Copy apps/server/.env.example to apps/server/.env and update PostgreSQL, Redis, and S3 credentials.
• Frontend: Copy apps/web/.env.example to apps/web/.env and set NUXT_PUBLIC_API_BASE_URL (default: http://localhost:9400).

3. Initialize Database Migrations

cd apps/server
bun run db:migrate

4. Launch Development Servers

From the repository root, start development servers concurrently:

bun run dev

---

🐳 Docker Compose Deployment (Single-Host / Self-Hosted)

Frontend and backend are merged into a single high-performance Docker image (Dockerfile). An embedded Caddy gateway transparently routes /api/* to the Hono backend and all other requests to the Nuxt frontend, exposing a single port (4000) with zero CORS issues:

# 1. Prepare environment file
cp .env.example .env
# Edit .env with your external DB_URL, REDIS_URL, S3 credentials, etc.

# 2. Build and start the unified container in the background
bun run docker:up   # Or: docker compose up -d

# 3. View logs
bun run docker:logs # Or: docker compose logs -f

Access endpoint:

• Stationary App (Web & API): http://localhost:4000

---

☸️ Kubernetes & GitOps (k3s + ArgoCD + GitHub Actions)

Stationary includes a complete GitOps deployment setup for Kubernetes / k3s:

• CI/CD: GitHub Actions (.github/workflows/deploy.yaml) builds and pushes the single unified image to GHCR (ghcr.io/stariverr/stationary) and updates the Kustomize image tag.
• GitOps: ArgoCD monitors deploy/k8s/overlays/production and automatically reconciles changes to the cluster.
• Ingress: Uses Traefik (default in k3s) to route all incoming traffic directly to port 4000 on the unified stationary service.

Apply the ArgoCD application to your cluster:

# 1. Create stationary secrets in k3s
kubectl create secret generic stationary-secrets -n stationary \
  --from-literal=DB_URL="postgresql://..." \
  --from-literal=REDIS_URL="redis://..." \
  --from-literal=AUTH_SECRET="..." \
  --from-literal=RESEND_API_KEY="..." \
  --from-literal=RESEND_EMAIL_SENDER="..." \
  --from-literal=S3_ENDPOINT="..." \
  --from-literal=S3_ACCESS_KEY_ID="..." \
  --from-literal=S3_SECRET_ACCESS_KEY="..." \
  --from-literal=S3_BUCKET="..." \
  --from-literal=CDN_BASE_URL="..."

# 2. Register ArgoCD Application
kubectl apply -f deploy/argocd/application.yaml

📚 Documentation

Detailed specifications and architectural guides are available in the docs/ directory:

• 📐 System Design & Database Specification - Data models, relational schemas, deletion policies, and JobRunner queue specs.
• 🔌 Metadata Sync & API Contracts - Integration client schemas, external_id deduplication rules, and sync payloads.
• 🔄 Metadata Ingestion Pipeline - Execution flow of saveMetadata, change detection, and S3 soft-delete lifecycle.
• 🔍 Multi-Space Hybrid Search Spec - Vector space segregation, pgvector HNSW indexing, and RRF rank fusion.
• ⚖️ Architectural Trade-offs - Design rationale behind Multi-Track schemas and DASH segment parsing.
• 🛠️ Task Engine Extension Guide - Strategy pattern guide for adding new background job handlers.
• 📜 TypeScript & Bun Guidelines - Mandatory standards for Temporal API, Web API streams, and bun:File.

---

📄 License

Private repository. All rights reserved.