The idea hits you in the parking lot. By the time you open Notes, pick a folder, and wonder if this belongs in Work or Random, the thought has already evaporated. You do not need another bloated knowledge base — you need a fast inbox for sparks, with search that actually finds them six months later.
Blinko (blinkospace/blinko) is an open-source, self-hosted card note app with optional AI-assisted retrieval (RAG over your own notes). ~11k GitHub stars, TypeScript stack, Markdown-friendly, Docker install in one line. Desktop builds use Tauri (macOS, Windows, Linux, Android). Think “Twitter-length capture meets personal wiki,” except the database lives on your server, not someone else's vector index.
What it actually does
Blinko is built for fleeting notes — short cards you jot quickly, tag, and link, not a 40-page Notion workspace. Plain text and Markdown keep friction low; the UI stays lightweight on purpose.
Where it differs from a static folder of .md files:
- AI-enhanced search — natural-language queries over your corpus via RAG, so “that hosting idea about Quebec latency” can surface a card you forgot you wrote.
- Self-hosted data plane — notes and embeddings stay in your deployment; you choose whether to call external LLM APIs or keep inference local.
- Multi-surface — Docker for homelab/VPS, Tauri apps for phone/desktop when you want offline capture that syncs back to your instance.
Official demo exists at blinko.space; production use is your own Docker host or desktop sync target. Read docs.blinko.space for env vars, auth, and backup paths before you expose it on the public internet.
Why self-host Blinko?
Notes are credentials-adjacent. People paste API keys, client names, half-baked pricing, and health stuff into quick captures. A SaaS note AI trains on aggregate behavior; Blinko keeps the corpus on disk you control — important for Canadian teams thinking about PIPEDA and “where does this text go when I search?”
RAG without renting memory. Managed AI note products charge per seat and per query. Self-hosting trades subscription for VPS RAM, disk, and the ops of patching Postgres (or whatever the compose file ships).
Speed of capture. Blinko optimizes for “open, type, done.” If your current stack is folders in git that you never search, a dedicated card UI with semantic search might actually get used.
Open source. Community on GitHub; Docker image blinkospace/blinko. Check license in repo before white-labeling for clients.
What running it takes
Quick start from upstream:
curl -s https://raw.githubusercontent.com/blinko-space/blinko/main/install.sh | bash(Repo org naming varies between blinkospace and blinko-space on badges — follow the install script on the branch you trust.)
Plan for: persistent Docker volumes, HTTPS reverse proxy, strong admin password, and backups of the database volume. AI features may need API keys for cloud models or extra CPU/RAM for local models — budget that before enabling RAG on a tiny VPS.
Compared to Wallabag (read-it-later) or enterprise wikis, Blinko is personal-scale capture plus search — not team permissions, not publishing workflows.
Who it's for (and who should skip it)
Good fit: developers and founders with idea churn, consultants logging client call crumbs, homelabbers who want AI search without uploading journals to a US SaaS, anyone replacing Apple Notes with something self-hosted.
Maybe skip it: teams needing shared workspaces and audit trails (use Outline, BookStack, or Notion). If you will not maintain Docker backups, a encrypted local app may be safer than a forgotten public Blinko instance.
Hosting it in Canada
We run small Docker stacks on Canadian VPS for personal knowledge tools — TLS, volume backups, and firewall rules so your note API is not world-readable.
Ask about sizing if you want Blinko beside other containers; the hard part is AI RAM and secure exposure, not the install script.
