Zhang Xiaofeng tomcatzh

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.

	uintmax_t parseHumanSize (const char* s) {
	char endp = (char )s;
	int sh;
	errno = 0;

	uintmax_t x = strtoumax(s, &endp, 10);
	if (errno \|\| endp == s) {
	errno = EINVAL;
	goto ERROR;
	}

	cue_file = 'file.cue'

	d = open(cue_file).read().splitlines()

	general = {}

	tracks = []

	current_file = None

	func NewGzipReader(source io.Reader) io.Reader {
	r, w := io.Pipe()
	go func() {
	defer w.Close()

	zip, err := gzip.NewWriterLevel(w, gzip.BestSpeed)
	defer zip.Close()
	if err != nil {
	w.CloseWithError(err)
	}

	#!/bin/bash

	if [ -z $1 ]; then
	FILTER='Name=instance-state-code,Values=16'
	else
	FILTER=$1
	fi


	IDs=`aws ec2 describe-instances --filters "$FILTER" \| jq '.Reservations[].Instances[].InstanceId' -r`

	#!/bin/sh

	yum update -y

	yum --enablerepo=epel install jq bash-completion -y
	cat <<EOT >> /etc/profile.d/aws-cli.sh
	if [ $SHELL = "/bin/bash" ]; then
	complete -C '/usr/bin/aws_completer' aws
	fi

	// Based on whitelist v1.2 by https://github.com/n0wa11

	function FindProxyForURL(url, host) {
	var PROXY = 'SOCKS 127.0.0.1:1983';
	if (isPlainHostName(host)) return 'DIRECT';
	if (/^\d+\.\d+\.\d+\.\d+$/g.test(host)) return 'DIRECT';
	var rules = [
	[
	'cn',
	'lan',

	func readLines(path string) ([]string, error) {
	file, err := os.Open(path)
	if err != nil {
	return nil, err
	}
	defer file.Close()

	var lines []string
	scanner := bufio.NewScanner(file)
	for scanner.Scan() {

	package main

	import (
	"encoding/json"
	"fmt"
	"time"
	)

	// The custom time format for json
	type MyTime time.Time