Pierre Romera Zhang pirhoo

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'm tired of extensions that automatically:
	// - show welcome pages / walkthroughs
	// - show release notes
	// - send telemetry
	// - recommend things
	//
	// This disables all of that stuff.
	// If you have more config, leave a comment so I can add it!!

	{

	https://www.facebook.com/sharer.php?u={url}
	https://www.facebook.com/dialog/share?app_id={app_id}&display={page_type}&href={url}&redirect_uri={redirect_url}
	https://reddit.com/submit?url={url}&title={title}
	https://twitter.com/intent/tweet?url={url}&text={title}&via={user_id}&hashtags={hash_tags}
	https://www.linkedin.com/sharing/share-offsite/?url={url}
	https://api.whatsapp.com/send?phone={phone_number}&text={title}%20{url}
	https://www.tumblr.com/widgets/share/tool?canonicalUrl={url}&title={title}&caption={text}&tags={hash_tags}
	http://pinterest.com/pin/create/button/?url={url}
	https://www.blogger.com/blog-this.g?u={url}&n={title}&t={text}
	https://www.evernote.com/clip.action?url={url}&title={title}

	{
	"vars": {
	"@gray-base": "#262626",
	"@gray-darker": "lighten(@gray-base, 13.5%)",
	"@gray-dark": "lighten(@gray-base, 20%)",
	"@gray": "lighten(@gray-base, 33.5%)",
	"@gray-light": "lighten(@gray-base, 46.7%)",
	"@gray-lighter": "lighten(@gray-base, 93.5%)",
	"@brand-primary": "#fedc32",
	"@brand-success": "#5cb85c",

	# Call scopes directly from your URL params:
	#
	# @products = Product.filter(params.slice(:status, :location, :starts_with))
	module Filterable
	extend ActiveSupport::Concern

	module ClassMethods

	# Call the class methods with names based on the keys in <tt>filtering_params</tt>
	# with their associated values. For example, "{ status: 'delayed' }" would call

	Step One - Starting with a single address field
	Edit Column > Add Column by Fetching URLs
	Nominatim has a limit of 1 geocode per second so make sure to set the throttle delay to greater than 1000 milliseconds
	Fetch URL based on column (quotes needed):

	"http://nominatim.openstreetmap.org/search?format=json&email=[YOUR_EMAIL_HERE].com&app=google-refine&q=' + escape(value, 'url')"

	------------------------------------------------------------------------------------

	Step Two - Extract lat/lon from newly created JSON blobs

	Services.factory("LawnchairFactory", function($window, $log, $parse) {
	return function(name, config) {
	var collection = {};
	var array = [];
	var isArray = config && config.isArray;
	var idGetter = $parse((config && config.entryKey) ? config.entryKey : "id");
	var transformSave = (config && config.transformSave) ? config.transformSave : angular.identity;
	var transformLoad = (config && config.transformLoad) ? config.transformLoad : angular.identity;

	function getEntryId(entry){

	/**
	* Patch the console methods in order to provide timestamp information
	*
	* Usage:
	* > console.log('ok')
	* 2012-09-06T11:52:56.769Z ok true
	*
	* Note:
	* The patch will only be applied with the first call.
	*

	sudo apt-get install openjdk-6-jre-headless
	curl -O http://dist.neo4j.org/neo4j-community-1.5.M02-unix.tar.gz
	tar -xf neo4j-community-1.5.M02-unix.tar.gz
	rm neo4j-community-1.5.M02-unix.tar.gz
	neo4j-community-1.5.M02/bin/neo4j start

	!!! 5
	html(class='no-js')
	head
	meta(charset='utf-8')
	meta(http-equiv='X-UA-Compatible', content='IE=edge')

	title

	meta(name='description', content='')
	meta(name='viewport', content='width=device-width, initial-scale=1')