Dax McDonald daxmc99

LLM Wiki

A pattern for building personal knowledge bases using LLMs.

This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.

The core idea

Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.

I get asked pretty regularly what my opinion is on merge commits vs rebasing vs squashing. I've typed up this response so many times that I've decided to just put it in a gist so I can reference it whenever it comes up again.

I use merge, squash, rebase all situationally. I believe they all have their merits but their usage depends on the context. I think anyone who says any particular strategy is the right answer 100% of the time is wrong, but I think there is considerable acceptable leeway in when you use each. What follows is my personal and professional opinion:

PostgreSQL is Enough

See the new site: https://postgresisenough.dev

Vagrant and VMWare Tech Preview 21H1 on Apple M1 Pro

UPDATE November 20, 2022: VMWare Fusion 13

VMWare Fusion 13 is now released. Read Vagrant and VMWare Fusion 13 Player on Apple M1 Pro for the latest.

Summary

This document summarizes notes taken while to make the VMWare Tech preview work on Apple M1 Pro, it originated

Helm 3 is storing description of it's releases in secrets. You can simply find them via

$ kubectl get secrets
NAME                                                TYPE                                  DATA   AGE
sh.helm.release.v1.wordpress.v1                     helm.sh/release.v1                    1      1h

If you want to get more info about the secret, you can try to describe the secret

$ kubectl describe secret sh.helm.release.v1.wordpress.v1

Instructions to hack up WSL2 on Windows 10 Build 18917 to run k3s (Kubernetes) and rio

Install WSL2

https://docs.microsoft.com/en-us/windows/wsl/wsl2-install

I already had Ubuntu-18.04 installed in wsl 1. So I just did wsl --set-version Ubuntu-18.04 2

Compile Kernel

Using Ubuntu 18.04 (I'm sure any distro will work), inside WSL2 download https://thirdpartysource.microsoft.com/download/Windows%20Subsystem%20for%20Linux%20v2/May%202019/WSLv2-Linux-Kernel-master.zip and extract to a folder. The latest version of the kernel source is available at (https://github.com/microsoft/WSL2-Linux-Kernel)

	# Required by intensive applications like ElasticSeach
	vm.max_map_count=262144

	vm.swappiness = 10

	vm.dirty_ratio = 80
	vm.dirty_background_ratio = 5
	vm.dirty_expire_centisecs = 12000

	alias dcup='docker-compose up'
	alias deploy='kubectl get deploy'
	alias dims='docker images'
	alias dps='docker ps -a'
	alias drm='docker rm'
	alias drmi='docker rmi'
	alias k='kubectl'
	alias ka='kubectl apply'
	alias kde='kubectl describe'
	alias kdl='kubectl delete'

	# Thanks to https://www.growingwiththeweb.com/2018/01/slow-nvm-init.html, modified
	# for zsh as "type -t" works in bash but not zsh

	# Add this to your .zshrc ...


	# Defer initialization of nvm until nvm, node or a node-dependent command is
	# run. Ensure this block is only run once if .bashrc gets sourced multiple times
	# by checking whether __init_nvm is a function.
	if [ -s "$HOME/.nvm/nvm.sh" ] && [ ! "$(type -w __init_nvm \| awk '{print $2}')" = function ]; then

	#!/usr/bin/env python
	import os
	import sys
	import subprocess

	if len(sys.argv) > 1:
	files = []
	for arg in sys.argv[1:]:
	if os.path.exists(arg):
	p = os.path.abspath(arg).replace('\\', '\\\\').replace('"', '\\"')