Selfhosted
A place to share alternatives to popular online services that can be self-hosted without giving up privacy or locking you into a service you don't control.
Rules:
-
Be civil: we're here to support and learn from one another. Insults won't be tolerated. Flame wars are frowned upon.
-
No spam.
-
Posts here are to be centered around self-hosting. Please ensure it is clear in your post how it relates to self-hosting.
-
Don't duplicate the full text of your blog or git here. Just post the link for folks to click.
-
Submission headline should match the article title.
-
No trolling.
-
Promotion posts require your active participation in selfhosting or related communities, or the post will be removed. No more than 10% of your posts or comments may be self-promotional, or your post will be removed. F/LOSS Exception: If your post is about a project that is completely open source & can be self-hosted in full without payment, and your account is at least 7 days old, your post is exempt from this rule as long as you continue to engage in comments.
Resources:
- selfh.st Newsletter and index of selfhosted software and apps
- awesome-selfhosted software
- awesome-sysadmin resources
- Self-Hosted Podcast from Jupiter Broadcasting
Any issues on the community? Report it using the report flag.
Questions? DM the mods!
view the rest of the comments
Qwen coder model from Huggingface, following the instructions there to run it in llama.cpp. Once that’s up: OpenCode and use the custom OpenAI API to connect it.
You’ll get far better results than trying to use other local options out of the box.
There may be better models potentially but I’ve found Qwen 2.5 etc to be pretty fantastic overall, and definitely a fine option beside Claude/ChatGPT/Gemini. I’ve tested the lot and it’s usually far more down to instruction and AGENTS.md instructions/layout than it is down to just the model.
Do you mind sharing your agents md?
This. Llama.cpp with Vulkan backend running in docker-compose, some Qwen3-Coder quantization from huggingface and pointing Opencode to that local setup with a OpenAI-compatible is working great for me.
The main thing that has stopped me from running models like this so far is VRAM. My server has a RTX 4060 with 8GB, and not sure that can reasonably run a model like this.
Edit:
This calculator seems pretty useful: https://apxml.com/tools/vram-calculator
According to this, I can run Qwen3 14B with 4B quant and 15-20% CPU/NVMe offloading and get 41 tokens / s. It seems 4B quant reduces accuracy by 5-15%.
The calculator even says I can run the flagship model with 100% NVMe offloading and get 4 tokens / s.
I didn’t realize NVMe offloading was even a thing and not sure if it actually is supported or works well in practice. If so, it’s a game changer.
Edit:
The llama.cpp docs do mention that models are memory mapped by default and loaded into memory as needed. Not sure if that means that a MoE model like qwen3 235b can run with 8GB of VRAM and 16GB of RAM, albeit at a speed that is an order of magnitude slower like the calculator suggests is possible.