During my early years as a computing enthusiast, Google Keep and Obsidian were the kingpins of my productivity arsenal. The former was fantastic for jotting down ideas and to-do lists, while the latter included all the features I’d need to build a knowledge base from the ground up.
However, Google Keep was the first to go after I began my descent into the self-hosting rabbit hole; its reliance on Google’s servers made it terrible from a privacy standpoint. Although Obsidian works really well online, I must admit that it added an extra layer of complexity to my setup. Luckily, I ran into Blink a few weeks ago, and it works as a solid alternative to both Google Keep and Obsidian once I paired it with my local LLMs.
How I use Dendron to organize project documentation across devices
Table of Contents
Dendron is an indispensable part of my coding stack
Even in a barebones setup, Blinko is a rock-solid note-taker and to-do list builder
I use “blinkos” to quickly jot down ideas
When it comes to tools bearing the AI-powered label (FOSS or otherwise), most of them either have gimmicky features or rely heavily on LLMs – to the point where they become useless without a clanker running in the background. Blinko manages to avoid these pitfalls, as it’s an amazing note-taker on its own. It supports three types of documents: notes, to-do lists, and “blinkos,” and each has its own unique functions. Blinkos, for example, are akin to flash notes taken as soon as inspiration strikes you. Rather than navigating through numerous folders and creating a new document, blinkos are meant to record random thoughts that you can ruminate on later.
I tend to use them all the time during meetings, Udemy lectures, and even while working on my server projects, and thanks to the random walk facility, I can review them at the end of the day. That’s where the notes section comes into the equation, as I can use it to expand upon the blinkos and turn them into proper documents I can cross-reference later. Then there’s the to-do list section, where I can create checklists for my everyday tasks.
The best part? All of them support typical Markdown syntax, so I can not only format them to my liking, but also add internal links to other notes and external websites. Blinko also lets me upload additional files and comments, and I can even check their revision history and create shareable links. Once I’m done processing my blinko ideas, I can swap them for the notes or checklist templates. This makes organizing my thoughts a lot simpler than transferring files between nested folders on typical note-takers.
On the surface, Blinko may seem somewhat minimalistic, and that’s its charm. It doesn’t overload me with tons of extra project management features, and instead provides a simple UI with some comprehensive note-taking facilities. But if you’re looking for some advanced tools, you can try arming it with the reasoning capabilities of LLMs…
LLMs can bolster its capabilities to the next level
Note analysis, tag generation, and RAG search bring it toe-to-toe with the best note-takers out there
Blinko may be a standalone app, but it can be paired with image, embedding, chat, and voice models to make it even more formidable. The auto tag creating facility is my favorite of the bunch, as it uses typical chat models to analyze my notes and come up with relevant tags for my notes. Likewise, every time I create a new blinko card, my LLMs add some extra context to it as a comment, and this comes in handy when I go through random meeting ideas over the weekend.
But the real game-changer is the retrieval-augmented generation facility, which lets me run queries on my notes and blinkos. While RAG is completely optional, feeding my notes into the LLM (and even random documents uploaded to the container) gives the AI models some reference material that they can ground their ramblings on, so I don’t get random incoherent results riddled with factual errors. Plus, it also helps me check the contents of note #1695 (and its associated resources) without going through them all.
I recommend pairing it with local LLMs
Ollama models work exceedingly well with this self-hosted app
Technically, Blinko supports Anthropic, Google AI, OpenAI, Azure, Grok, and a bunch of other cloud providers for the LLM-powered features. But considering the privacy advantages of this self-hosted app, I’d rather not tarnish them by throwing external AI providers into the mix. As such, I use Ollama models running on a nearly decade-old Pascal card with Blinko, and this setup works surprisingly well.
First, I deployed Blinko by running the curl -s https://raw.githubusercontent.com/blinko-space/blinko/main/install.sh | bash command in a Linux terminal. Then, I used the IP address of the VM, followed by a colon (:) and port number 1111 to launch Blinko’s web UI. Once I’d created a new user, I headed straight to the Settings page and connected the Ollama LXC bearing all my models with Blinko. Deepseek R1 (7B) is better for summarizing my notes, while Qwen3 (8B) and Llama 3.1 (8B) are good at creating tags. For the RAG tasks, I use nomic-embed-text as the embedding model.
Meanwhile, voice models only work with OpenAI models, so I couldn’t get them running on Blinko. Likewise, I had trouble using the online search facility when chatting with my AIs, as it relies on a Tavity search API key, and I refuse to bring external services into the fray. But considering that I use Open WebUI + SearXNG if I need to browse the web for research, I have zero issues with relying on the Blinko + Ollama setup for my note-taking needs.
