Aether — Personal Knowledge Engine

Where links become knowledge

Save any URL — YouTube, Instagram, Twitter/X, articles, PDFs — and AI extracts, transcribes, summarizes, and organizes the content into a searchable, multi-vault knowledge base with an AI chatbot assistant, knowledge graph, and entity extraction.

┌─────────────┐     ┌─────────────┐     ┌─────────────┐
│   Frontend   │────▶│   Backend   │────▶│   Worker    │
│ React/Capacitor│   │   Go/Gin    │     │  Python/    │
│ iOS/Android/Web│   │   :8080     │     │  Celery     │
└─────────────┘     └──────┬──────┘     └──────┬──────┘
       │                   │                    │
  ┌────┴────────┐   ┌──────┴──────┐      ┌──────┴──────┐
  │ iOS App     │   │  PostgreSQL │      │    Redis    │
  │ Android App │   │  (pgvector) │      │   (queue)   │
  │ Chrome Ext  │   └─────────────┘      └─────────────┘
  └─────────────┘

• Multi-Vault — Organize notes into separate vaults; all content (notes, labels, entities, relations) is vault-scoped via the X-Vault-Id header
• Multi-platform URL capture — iOS Share Extension, Android Share Intent, Chrome Extension, Web
• AI Processing Pipeline — YouTube transcription, Instagram OCR, Twitter/X scraping, PDF extraction, article parsing
• AI Summaries — Claude CLI generates structured Markdown with entity extraction, Gemini fallback
• Community Insights — YouTube/Instagram/Twitter comments analyzed by AI
• Semantic Search — pgvector 384-dim embeddings for intelligent search

• Entity Extraction — AI extracts people, concepts, tools, books, films, locations, etc. from every note
• Entity Graph — Entities become hub nodes connecting notes that mention the same person/concept/tool
• Entity Browser — Browse all entities by type, search, click through to related notes
• Cross-Reference & Note Linking — pgvector similarity finds related notes, creates bidirectional links
• Knowledge Graph — Interactive force-directed visualization with dual mode (Similarity / Entities)
• Activity Log — Tracks note processing, relation discovery, entity extraction
• AI Rules — Custom user instructions for how AI processes content

• Aether AI Chatbot — Vault-aware chatbot powered by Gemini 2.5 Flash
• SSE Streaming — Real-time token-by-token response streaming
• Note Linking — AI references specific notes with clickable links
• Conversation History — Persistent chat sessions grouped by date
• Save to Vault — Save chat insights as new notes

• Smart Topic Labels — AI auto-generates 2-4 topic labels per note (i18n aware)
• AI-Generated Titles — Clean, descriptive titles from AI analysis
• Source Labels — Auto-labels by domain (YouTube, Instagram, Twitter/X, etc.)
• Label Filtering — Filter vault by label, with translated label names

• Knowledge Dashboard — Stats, topic distribution, entities, recent notes, activity log
• Onboarding — 7-screen flow (Welcome, Language, AI, Assistant, Privacy, Tutorial, Notifications)
• Splash Screen — Code-generated halftone dot pattern animation
• Infinite Scroll — Paginated vault (20 notes per page)
• Compact List View — Alternative list layout with thumbnails and labels
• Glassmorphism Tab Bar — Blur effect bottom navigation
• Neon Heartbeat FAB — Animated AI chatbot button with pulse rings
• Notification Center — Full-page notification list with history
• Dark Obsidian Theme — Premium dark UI with purple accents

• Admin Panel — Owner-only (ADMIN_EMAIL) dashboard at /admin: all users, what they've saved, global stats & recent feed (read-only)
• System Monitoring — scripts/aether-monitor.sh (cron) → ntfy alerts on container down, disk full, Redis/Postgres down, worker wedge, stuck notes, error spikes
• Sign in with Apple + Google — Both providers on iOS/web
• Account Deletion — Full GDPR-compliant data removal
• Push Notifications — FCM server-side push when processing completes
• Bilingual — Full English/Turkish support (UI + AI + labels)
• Firebase Analytics — 14 custom events, screen tracking, platform detection

• Theme: Dark Obsidian (#0e0e0e) with purple accent (#b79fff)
• Logo: Halftone dot pattern — circular grid with radial size gradient
• Fonts: Space Grotesk (headlines/labels), Manrope (body)
• Design Tool: Google Stitch (MCP integrated)

aether/
├── backend/          # Go API server
│   ├── handlers/     # notes, users, vaults, chat, search, labels, graph, entities, activity, sse, desktop_auth
│   ├── middleware/    # Firebase auth + VaultResolver (X-Vault-Id)
│   ├── models/       # User, Vault, Note, NoteRevision, Label, Entity, NoteEntity, NoteRelation, ChatMessage, ActivityLog
│   ├── config/       # Environment config
│   └── database/     # DB connection & migrations
├── frontend/         # React SPA + Capacitor
│   ├── src/
│   │   ├── components/   # Sidebar, AetherChat, SplashScreen, LabelManager, VaultManager, VaultSwitcher, MobileVaultSheet, editor/ (TipTap)
│   │   ├── pages/        # Dashboard, Vault, Editor, Graph, Entities, EntityDetail, Chat, Activity, DesktopAuth, Onboarding, Settings, Login
│   │   ├── contexts/     # Auth, Language, Vault providers
│   │   ├── i18n/         # en.js, tr.js translations
│   │   ├── analytics.js  # Firebase Analytics
│   │   └── api.js        # API client with SSE streaming
│   ├── src-tauri/        # Tauri 2.0 desktop app (macOS/Windows/Linux)
│   │   ├── src/          # Rust backend (menus, plugins, setup)
│   │   ├── capabilities/ # Tauri permissions
│   │   ├── icons/        # Desktop app icons
│   │   └── tauri.conf.json
│   ├── ios/App/          # Xcode project + AetherShare extension
│   └── android/          # Android Studio project
├── worker/           # Python Celery worker
│   ├── tasks.py      # URL extraction, AI processing, entity extraction, FCM push
│   └── celery_app.py # Celery configuration + Beat schedule
├── extension/        # Chrome extension (MV3)
├── landing/          # Static landing page
└── docker-compose.yml

POST /api/v1/share → Create note (status: processing) → Redis → Celery worker
  1. extract_content_from_url()     — YouTube/Instagram/Twitter/article extraction
  2. call_llm()                     — Claude CLI (primary) or Gemini (fallback)
  3. extract_comments()             — YouTube/Instagram/Twitter comments
  4. generate_community_insights()  — AI analysis of community discussion
  5. generate_title()               — Clean AI-generated title
  6. update note status → ready
  7. generate_embedding()           — sentence-transformers → pgvector
  8. auto_label_source()            — Label by source domain
  9. auto_label_topics()            — AI extracts 2-4 topic labels
 10. find_related_notes()           — pgvector similarity → note_relations (powers Graph)
 11. extract_entities()             — AI extracts people, concepts, tools, etc. (powers Entities)
 12. send_push_notification()       — FCM push to user

Server: Self-hosted Linux (Docker Compose) via Tailscale
Proxy: Nginx Proxy Manager
CI/CD: GitHub Actions → rsync → docker compose build/restart

# Server deploy
rsync -avz backend/ root@minis:/root/aether/backend/
rsync -avz worker/ root@minis:/root/aether/worker/
rsync -avz frontend/ root@minis:/root/aether/frontend/
ssh root@minis "cd /root/aether && docker compose build api worker frontend && docker compose up -d api worker frontend"

# Desktop release (triggers GitHub Actions)
git tag v1.x.x && git push origin v1.x.x
# → Builds macOS (ARM + Intel), Linux (deb + AppImage), Windows (MSI + EXE)
# → Creates GitHub Release with all artifacts

# Local desktop build
cd frontend && npm run build && npm run build:desktop
# → Output: src-tauri/target/release/bundle/dmg/Aether_*.dmg

Private project.