Leo Dutra leodutra

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.

JavaScript Bitwise Hacks

It assumes the highest positive signed 32-bit float value for numbers. In other words, 2147483647 (or 0x7FFFFFFF or 2^31-1).

News

Galnet (in-game)

Return the Galnet news.

Base URL: https://cms.zaonce.net/en-GB/jsonapi/node/galnet_article

You can change the en-GB in the URL to get Galnet in others languages (fr-FR, pt-BR...).

Ubuntu

Mount rclone sync

This config allows us to mount obsidian folder locally, preventing issues with Google Drive hashes

# create obsidian folder on Google Drive, then
# install rclone
curl https://rclone.org/install.sh | sudo bash

	sudo pacman -S --needed base-devel git pkgconf cmake make gcc

	# Install Paru
	if ! command -v paru >/dev/null 2>&1; then
	git clone https://aur.archlinux.org/paru.git
	cd paru \|\| exit
	makepkg -si
	fi

	# Install Fabric AI + dependencies

	#!/usr/bin/env bash
	set -euo pipefail

	# ============================================================
	# Rust development environment bootstrap (Arch Linux + zsh)
	# ============================================================

	# ------------------------------------------------------------
	# Helpers
	# ------------------------------------------------------------

	#!/usr/bin/env bash

	# sets MAKEFLAGS in /etc/makepkg.conf
	# so makepkg (AUR/package builds) uses all CPU cores minus one for faster compilation
	sudo bash -c 'grep -q "^MAKEFLAGS=" /etc/makepkg.conf && sed -i "/^MAKEFLAGS=/c\MAKEFLAGS=\"-j$(( $(nproc) - 1 ))\"" /etc/makepkg.conf \|\| echo "MAKEFLAGS=\"-j$(( $(nproc) - 1 ))\"" >> /etc/makepkg.conf'

	# ------------------------------------------------------------
	# Helpers
	# ------------------------------------------------------------
	zshrc_append() {

	"use strict";

	/**
	* Hypertext Transfer Protocol (HTTP) response status codes.
	* @see {@link https://en.wikipedia.org/wiki/List_of_HTTP_status_codes}
	*/
	enum HttpStatusCode {

	/**
	* The server has received the request headers and the client should proceed to send the request body

	import { type RefObject, useEffect, useState } from 'react';

	export type IntersectionObserverOptions = IntersectionObserverInit & {
	intersectOnce?: boolean;
	skip?: boolean;
	};

	const emptyOptions: IntersectionObserverInit = {};

	const useIntersectionObserver = (

	height: 700
	scrolling: yes