FUD

Gitea is still MIT licensed and the Enterprise tier features only cater to large org needs [1]. Why would I want to deal with nightmarish SAML config when OIDC does a better job

Forgejo was forked because the maintainers were butthurt they didn’t have more say in the development roadmap and their large PRs didn’t get reviewed and merged fast enough. Which is a valid reason to fork

I wish more people understood this.

Do I use LLMs to write software for my personal use? Sure. I still try to build it in an “incremental” way the same way I would write software manually so I don’t get 10k SLOC written in a week, but at some point, even reviewing 100 LOC changes takes time, so I just take a cursory look at the diff, yolo-merge-and-run to test it. It’s not critical. This is fine. I’m just exploring the problem domain and solutions.

But would I go as far as sharing it, making a damn git repo and advertising it on Lemmy? Fuck no. This is unreliable, inscrutable slopware tailored for my own use. Anyone with a local LLM or a 20 euro claude/codex/z.ai subscription can do the same thing in a few minutes of work a day.

A single 10-line patch/contribution to a human-written project, with contributors who understand the code, or even a well-curated comment in a bug tracker that helps devs debug an issue or clearly expresses a need, has more value for the community than 50 vibe-coded projects.

I didn’t find one that would work entirely authelssly and which would allow negative entries

Have you considered filing feature requests for these on existing, well-maintained projects?

Yes, even writing a proper issue report probably entails more work and brainstorming than prompting your way to a shitty solution. No offense meant, I do it as well. I know using a LLM and pumping out a working solution to a complex problem in a week gives a feeling of euphoria and power; this wouldn’t have been possible at all a few years ago. But there is absolutely no value in proactively sharing and advertising it.

I appreciate OP being transparent about it though.

Yep. https://lemmy.world/post/46066942/23416719 Basic setup, works for me

they don’t go out of their way to make the self hosted option easy

Just follow the docs https://docs.nextcloud.com/server/stable/admin_manual/installation/source_installation.html ? The manual install is a simple webserver + PHP-FPM + postgres setup

I have automated it with an ansible role [1], there’s nothing complicated about it, really.

That’s only if you use the default Build agent with the built-in prompt (https://github.com/anomalyco/opencode/blob/dev/packages/opencode/src/session/prompt/default.txt), and yes it is quite large.

It’s trivial to create custom agents in opencode.json with custom prompts, tools, whatever…

For example I have created a Personal agent which handles menial stuff such as searching/editing my notes, appointments, tasks… with a restricted set of tools and skills.

The single most important change I made is only allowing the local provider in the config, which disables all cloud providers. IMHO this should be the default but I’m not complaining. It’s the best open-source harness I’ve tried so far. I want to try pi.dev someday (quite minimal, needs a good amount of setup and tuning).

I also argue that some local models actually behave much better with a semi-large system prompt (qwen 3.6 for example tends to lose itself in reasoning if you only use the default You are a helpful assistant system prompt and a basic Say hi user prompt - opencode-like large system prompts fixes this; even if you lose some time for initial prompt loading)

I suggest using llama.cpp instead of ollama, you can easily squeeze +10% in inference speed and other memory optimizations from llama.cpp. With hardware prices nowadays I think every % saved on resources matters. Here is a simple ansible role to setup llama.cpp, it should give you a good idea of how to deploy it.

A dedicated inference rig is not gonna be cheap. What I did, since I need a gaming rig; is getting 32GB DDR5 (this was before the current RAMpocalypse, if I had known I would have bought 64) and an AMD 9070 (16GB VRAM - again if I had known how crazy prices would get I’d probably ahve bought a 24GB VRAM card). The home server runs the usual/non-AI stuff, and llamacpp runs on the gaming desktop (the home server just has a proxy to it). Yeah the gaming desktop has to be powered up when I want to run inference, this is my main desktop so it’s powered on most of the time, no big deal

See you back on Debian in a few months