Chris Shabsin's personal page for tracking projects and such.
This document tracks the setup of my new AI development machine, “cricket”.
python3, python3-pip, and python3-venv./home/cshabsin/proj/ComfyUI.cvdl alias from the old Windows PC to download models from Civitai with an API key automatically.ComfyUI/models subdirectories.To run:
$ . proj/venv/bin/activate
$ cd proj/ComfyUI
$ python main.py --listen
nvm (Node Version Manager) using the curl | bash method in WSL.nvm to install the Gemini CLI.Installed koboldcpp and SillyTavern. koboldcpp is a downloaded binary in proj/koboldcpp, and SillyTavern is a git clone in proj/SillyTavern.
To run:
koboldcpp from its directory and pick a .gguf file to load../start.sh --listen in the SillyTavern directory.The firewall is set up to allow port 8000 through to Tailscale, and it can be accessed with http://cricket.dog-garibaldi.ts.net:8000 from within the Tailscale private network.
Installed Ollama in WSL using the “curl” method from their website and used it with the continue.dev VS Code extension.
There is a known conflict between the Node.js version used for the Gemini CLI and the version required for the conjunew project.
nvm use 24).conjunew project: Requires Node.js v22 due to Firebase Functions compatibility (nvm use 22).Use nvm to switch between these versions when working on the respective projects.
These require Java, so we install the openjdk-21-jdk Ubuntu package, then we can install the firebase tools:
$ sudo apt-get install openjdk-21-jdk
$ npm install -g firebase-tools
GEMINI.md files from other systems and synthesize the ideas.ssh-agent for improved SSH key management.