Kev's Development Toolbox

I want to use a Qwen 3.6 local model but have found from testing with various harnesses it is prone to overthinking and goes in circles on some tasks.

Several months back I updated on my progress on moving a number of my personal projects from botsin.space mastodon server to my own https://mastodon.kevinhooke.com instance.

For running local LLMs, ollama is an easy way to get up and running, but if you need or want to start tuning model parameters, llama.cpp offers more flexibility (and is generally faster than ollama) I installed following steps here.

July 2026: I have a desktop with an Nvidia 2060 with 6GB and was curious whether I’d get any usable performance if I attempted to run an LLM model locally What have I tried so far: Update Aug 2026: since I started writing this I’ve upgraded to a 5070ti wi…

Create a mount point: sudo mkdir /mnt/[mountpoint] To temporarily mount the share: sudo mount -t nfs [nas-ip]:/volume1/[nfs-sharename] /mnt/[your-mount-point] To mount the share permanently, add to /etc/fstab and then sudo mount -a: [nas-ip]:/volume1/[nfs…

I’m running Ollama as a service on Ubuntu 24.04, and after starting openweb-ui as a Docker container with: docker run -d -p 3000:8080 --network=host -v open-webui:/app/backend/data -e OLLAMA_BASE_URL=http://127.0.0.1:11434 --runtime=nvidia --name open-web…

If you follow a ‘normal’ install of docker via the install guide, you add yourself to the docker group to get permissions as a normal user to run the docker commands.

by default, ollama sets a very low context size with any model it launches.

I worked on a project where there was a ‘bug bounty’ with prizes for fixing the most bugs in a month.

Assuming you have ollama installed and a local model already downloaded, install opencode following steps here.

docker stop [portainer-container-id]docker run --rm -v portainer_data:/data portainer/helper-reset-password