13
votes
What programming/technical projects have you been working on?
This is a recurring post to discuss programming or other technical projects that we've been working on. Tell us about one of your recent projects, either at work or personal projects. What's interesting about it? Are you having trouble with anything?
I've been obsessively working on my Notelog project. Notelog is a CLI note-taking program that also has a Model Context Protocol server built-in, so that it can be used by AI assistants.
I'm now done with all the functionality I imagined for the tool -- note creation, note searching (via SQLite's FTS5 full-text search), and editing of tags works by just asking the LLM. I'm using Gemini 2.5 Flash via Dive -- Dive is not super polished, but there does not appear to be another MCP client that can both live in my system tray and works under Linux. It's nice to be able to just chuck a "/log this-or-that" line at Gemini, and have it make a grammatically correct note of it, with automatically selected title and a few tags (though it doesn't do that 100% reliably, which is a bit annoying).
The project now clocks in at about 5000 lines of code (mostly AI-written, though I do read and fix the code if there are issues). I think it's an example of software that is uniquely enabled by AI-coding, because writing everything myself would have taken so long that I would have abandoned the project before it got to a useful state. It's also software that is somewhat unlikely to be useful for anyone but me... I expect a lot more software that is only useful to a small group of people or single person will get written as prices go down due to LLMs getting better at writing software fast.
One gimmick I started experimenting with is giving the AI access to my self-hosted Forgejo instance. I made a user for it, created an access token, and now it can read and comment on issues in projects I invite it to. Instead of writing a long prompt inside of the cramped IDE, I can just create (and preview!) an issue and then tell it to read and implement that -- a kind of poor man's Codex, if you will.
Migrating my self-hosted environment, consisting of a bunch of containers managed by docker-compose, to kubernetes.
The plex stack (including sonarr, radarr, prowlarr and qbittorrent) was easy to migrate.
My homeassitant ecosystem (including zwavetomqtt, zigbeetomqtt, mosquitto, some voice assist AI components, etc) was a lot more tricky.
I wrapped things up with some utility containers, samba, nfs, and some cronjob containers. Finally also added wireguard as a vpn, also in container format.
All this is working with certmanager and nginx as ingresses to automate certificates and DNS.
One last thing I want to look into is network isolation between the components using cillium.
What prompted you to do that?
I currently have a hacky setup using systemd services + random shell and python scripts, without any higher-level management.
I plan to migrate this to a local server after my exam period, and I planned to do a podman-compose setup, which sounds much like what you had before.
It would be interesting to hear what were you dissatisfied about with your previous setup (or whether it was just for fun to learn & play with new tech, which is a big part of my upcoming migration).
Hi, thanks for asking! No dissatisfaction! The compose setup was working just fine.
However I have wanted to get into k8s for a little while now and I personally believe in diving straight in with a real project. So the plex and homeassistant ecosystems were perfect for this.
I must say though, one big advantage is the level of automation regarding certificates and DNS.
This is very much possible without k8s but it really works so seamlessly using certmanager.
For a couple of months I've been slowly noodling in my free time on a programming environment where you pervasively have the ability to draw boxes and arrows and create hyperlinks from one part of the source code to another.
https://akkartik.name/post/2025-03-08-devlog
https://merveilles.town/@akkartik/114547652849162554
Writing a Swift iOS app to help myself learn how to deal with being autistic with an eye toward making a SaaS out of it if I'm successful.
If it scratches my itch, it may well help others.
Implemented editing via web interface for the (non-technical) team on our internal wiki. This required improving deployment to our on-prem Synology. Previously I was using a half-baked script on my machine that just SSHed in and pulled the latest changes from origin (I was the only one editing the wiki).
Now it’s a bit more solid — a bare repo with an external worktree, the Synology as external remote, and a nightly script that commits and pushes web edits back to origin. Nothing fancy, but fun to set up.
The wiki is a customized, self-hosted version of w2wiki, which is pretty phenomenal. Simple, small, and easily hackable.
My Fedora installation ended up having some issues with BTRFS where I appear to be missing some checksums, so currently been researching that. I was able to do a live boot, and have since backed up all my documents on my NAS (I should really automate some back ups). Weirdly, my Windows partition (on the same drive) was also having issues at the same time. Although, it had been probably a month or two since Windows was last used so not sure if the issues happened at the same time. Windows was able to self-repair itself, so now I am researching BTRFS some more to try and see if I can fix the issue, but if not I will just reinstall my Fedora partition.