Personal note-taking app

I wanted a easy-to-use self-hosted note-taking app that I am able to pick up, throw ideas at, and let a simple local LLM do the boring work of categorizing, tagging, and fetching notes.

I think one aspect that is often overlooked these days with AI is that there is a lot of cool things we can do with lightweight local models. Such as tagging posts with pre-defined categories, adding metadata, embedding text and images, performing sentiment analysis etc. Automating the boring stuff. Filling in categories for blog posts or well thought-out pieces of writing is fine but what if you have to do it 30 times in a day?

I recently got back into reading more non-fiction, and one of the things I miss is a convenient way to take notes. For physical notes I need to always carry a notebook, so I prefer my phone, which can also accept and transcribe voice-memos. But then what? Create a new note with iOS note taking app, or the Nextcloud notes? Sure, but if I make many such notes it’s a bit of a pain to tag and categorize all of them. Also I prefer not using closed source note-taking software whenever possible. And Nextcloud notes are convenient but only have the most bare bones features.

So my only choice is building something myself, and maybe I’ll even learn some things on the way.

I decided to build Bunko (“book collection” in japanese) with the help of Deepseek V4 flash 1017 with the Pi harness running sandboxed on my Homelab raspberry pi. And since the same raspberry pi is on my tailnet, iterating becomes quite easy. I just ask the agent to deploy it locally on the raspberry pi, which allows me to access from anywhere else within the same tailnet.

I also had the opportunity to try herdr with a remote session, which worked flawlessly.

You can find the code here on my github: Bunko code