Rene Dworschak rddaz2013

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.

	#!/bin/bash
	# TODO: skip tiny files (so small they couldn't be photos)
	# TODO: make sure sym links and other file system oddities are handled
	# TODO: look at paralellization for perf boost

	#
	# Constants
	#
	CHAR_COUNT=12
	BLOCK_COUNT=6

	/* SETTINGS */
	var l='';
	var p='';
	var area = 'https://www.ingress.com/intel?ll=35.5841133,139.4988137&z=14';
	var v = 20000; //Delay between capturing screenshots, in milliseconds (1000 ms = 1 s)
	var reloadtime = 60000 * 120;
	var width = 900; //Picture width
	var height = 500; //Picture height
	var folder = './'; //Folder where to save screenshots, with / (or \) on the end. '.' means current folder.
	phantom.injectJs('settings.js');

	from flask import Flask, send_file, send_from_directory
	from watchdog.observers import Observer
	from watchdog.events import FileSystemEventHandler
	import os

	NBCONVERT_PATH = '../nbconvert/nbconvert.py'
	TARGET_IPYNB = 'presentation.ipynb'
	TARGET_HTML = 'presentation_slides.html'

	def nbconvert():